Bitcoin Forum

The mailman brought my brand new Radeon HD 7970 today and of course one of the first things that I wanted to try with it was to see how well it can mine :)

I've noticed that there's an almost complete lack of benchmarks for this card on the web (the only one I've seen has been the 414MH/s result from tom's hardware) so I've created an account to share my results here.

I'm using cgminer-2.1.1-win32 as my mining software and figured that this would be a plug-and-play deal, but I ran into a couple of initial snags. The first thing I noticed is that the card would report that it was hashing but would never get any accepted results. I was unable to get the phatk kernel working, however I found that by using the "--kernel poclbm" option and editing line 36 of poclbm110817.cl the following way to disable BFI_INT that I would start getting accepted hashes:


With those two changes to the default configuration of cgminer hashes start to get accepted, but the 290MH/s hashing performance with the default settings (-g 2 -v 2 -w 128) for this kernel were slower than the 310MH/s from the trusty OC'd HD5850 that this new card replaced, so I played around with the --gpu-threads, --vectors and --worksize settings and here's a small table with the results:

--gpu-threads 1 --vectors 1 --worksize  32 : 224MH/s
--gpu-threads 1 --vectors 1 --worksize  64 : 448MH/s
--gpu-threads 1 --vectors 1 --worksize 128 : 446MH/s
--gpu-threads 1 --vectors 1 --worksize 256 : 448MH/s

--gpu-threads 1 --vectors 2 --worksize  32 : 141MH/s
--gpu-threads 1 --vectors 2 --worksize  64 : 285MH/s
--gpu-threads 1 --vectors 2 --worksize 128 : 283MH/s
--gpu-threads 1 --vectors 2 --worksize 256 : 284MH/s

--gpu-threads 1 --vectors 4 --worksize  32 :  66MH/s
--gpu-threads 1 --vectors 4 --worksize  64 : 133MH/s
--gpu-threads 1 --vectors 4 --worksize 128 : 133MH/s
--gpu-threads 1 --vectors 4 --worksize 256 : 133MH/s

(Note: all of these results were using the standard 925MHz core and 1375Mhz mem clocks)


Given these results I figured that -v 1 -w 64 was probably the best kernel setting so I tweaked the number of threads:

--gpu-threads 1 --vectors 1 --worksize  64 : 448MH/s
--gpu-threads 2 --vectors 1 --worksize  64 : 467MH/s
--gpu-threads 3 --vectors 1 --worksize  64 : 474MH/s
--gpu-threads 4 --vectors 1 --worksize  64 : 472MH/s
--gpu-threads 5 --vectors 1 --worksize  64 : 473MH/s
.
. (No apparent trend so I skipped these...)
.
--gpu-threads 8 --vectors 1 --worksize  64 : 473MH/s


So the best settings I could come up with for the day were -g 3 -v 1 -w 64 giving a hashing rate of 474MH/s! Not too shabby considering the card is using stock speeds! And its roughly a 15% higher hashrate than what was reported at tom's.

I'll continue experimenting with other settings over the next few days, but this initial testing makes me wonder what a kernel rewritten specifically for GCN (using the 16-wide SIMD, among other things) could do.


**** UPDATE ****

Someone suggested that I give a recent version of the DiabloMiner a try since it should have decent support for GCN, so I did.

~650MH/s with the default diablominer settings and the card OC'd @ 1125/975MHz:

https://i.imgur.com/f8NnZ.jpg (http://imgur.com/f8NnZ)

~530MH/s at standard clocks:

https://i.imgur.com/DeUxm.jpg (http://imgur.com/DeUxm)

How did you acquire an HD 7970?

Interesting first post.

Thanks for that.  Where'd you get the card from?

474 is already nice!

And I'm sure that GCN optimized kernel should pass 500 :)

Thanks! I did a google shopping search for the card on a whim and surprisingly found it in stock at a nice on-line store in Spain called pccomponentes.com

Ordered it on Jan 2nd, and it arrived on the 4th.

(Edited dates. Didn't notice it was past midnight already :)

Nice score for you.  Utterly stupid for "pc componentes".  AMD generally frowns on early releases.  Likely the carton had a giant red tape across it saying "not for retail distribution until January 9th".  Maybe it wasn't in Spanish?  :)

Still pretty sweet.

The card has been running stable for 4hours now and averaged 467MH/s at stock speeds over that timeframe.

Just took a stab at overclocking the card and it pushes an average of 568MH/s @ 1125MHz. The overdrive panel won't go any higher so that's it for the time being. The stock cooler keeps the card at around 81C in my PC with automatic fan speeds.

I already said that somewhere around the forum, the 7970 will kick some serious ass!
And let's wait for GCN optimizations...

81°C at 1125? That's awesome.
Any chance of you measuring the card's power draw? (If you have a kill-a-watt just measure the machine's power draw in idle and during mining. The delta is the 7970's power draw).

Total BS. We are not dumb. You work for AMD. They have finally waken up to this slice of the market. Just wait and see what Nvidia got in store for later on probably when they also wake up as well. Post the link on the exact site if you are not selling us BS. Who gets a card that is going to be released in 9 Jan today ? Newbie account too. LOL. AMD PR fail just like their drivers and QC. We had the CPU bug for ages. If it wasn't for the integer performance I would have gone to the green side long ago. Hopefully they got a Nvidia mining card in stock too at that pccomponentes.com fake site. Searching for 7970 brings up NOTHING.

Power figures are essential and prices. Sucks still if the cost and power is too damn high. 5XXX still the best for now at US prices.

Somebody needs to ring the damn bell over at N-"we-don't-care-about-mining"-vidia. Needing to have the X server active to mine = fail.

Here's what I read:
Blablabla drama blablabla I MAD blablablabla AMD FAIL blabla I JELLY blablabla GO NVIDIA blablabla.

I don't like being tricked. This is clearly an AMD rep trying to pass off as a casual miner. This is deception and I don't take very kindly to that.

Now they suddenly wake up and include mining as a feature but until now we had to put up with crap drivers and CPU bugs etc.

Sounds like a lot of hypocrisy going on at AMD HQ. Hopefully Nvidia wakes up to bring some competition into this game because I am sick of AMD and their crap and failing fans and all that LQ BS.

Competition and Nvidia coming on board would be really excellent. I am sick of putting up with AMD's BS just because they are the only player in the GPU market. I hope Nvidia make a dedicated folding / mining card without display output and all those useless components and really owns AMD and teaches them not to be complacent in regards to mining, drivers, CPU bugs and all the crap they could do till now ( if Nvidia hopefully also wakes up ). 

LOL. Gotta admit they tricked me with this one. I really believed it. Probably a hoax with all the people saying 7970 and CGN suck at mining. Good one !

If you can use a kill-a-watt and measure power at the wall.  If you measure idle, load, load overclocked that will give us the load of card @ stock and overclocked.

Quick google shopping search led me to: http://www.excaliberpc.com/611611/sapphire-radeon-hd-7970-3gb.html

$647.22 (with shipping) for 568Mh/s doesn't get me very excited. Even when optimize IMHO you will most likey be paying $1/Mh just for the cards while they are in their release stage.

Edit: I would also like to hear power draw numbers.

I'm afraid I don't have a kill-a-watt, but if it helps the power draw has been pretty well tested on the review sites and I assume that my card is running within the same ballpark.

I've done some further testing with the overdrive panel and managed to underclock the memory to 950Mhz which hopefully will lower the power draw and temps. The slider in the interface claims to go all the way down to 150Mhz but I can't seem to get it to recognize settings any lower than 950Mhz.

Something else I noticed is that the base plate gets quite hot during mining. I'm not sure whether its because the temperature sensor on the 7970 is being optimistic with its 81C reading or if the new heatsink design is doing a great job at spreading heat but the base plate is a lot hotter to the touch than my 5850 (which would only get hot near the VRMs). It makes me kind of worried that the VRMs or memory might be getting too hot so until GPU-Z starts to report other temperature sensors apart from just the GPU sensor I'm going to stick with manually setting the fan speed from the overdrive panel. Currently I have it at 60% speed giving an average temp of 69C @ 1125/950Mhz.

I understand all the skepticism about these results. Believe it or not I don't work for AMD or any other PR firm. I also wish that NVidia would crank out better numbers considering that my second miner is a workstation with a Quadro 5000 series that pushes 55MH/s.

When I ordered the HD 7970 I also thought that it would arrive on the 9th at the earliest and couldn't believe it when I got a message that it was in transit the next day. I got lucky I guess. The card has been de-listed from the website since, I don't know if its just out of stock because of "limited availability" or if the AMD marketing dept made a call. However doing google shopping searches shows at least one other vendor in my area with the card.

And this is just a guess, but I think that the difficulty in actually getting the mining software to work with this card is what turned many review sites off of actually posting any bitcoin hashing figures. With cgminer's default settings the performance is lackluster and unless you edit the OpenCL source like I did hashes aren't even accepted, so reviewers probably just chalked it off as software/driver bugs and moved on to other, working, benchmarks in the race to post their reviews.

Hopefully someone else with a HD 7970 card who was having trouble getting the card to mine can apply my tweaks and reproduce the results on their end?

Eventually.  List price is $549 and is looks like availability will be tight so expect retailers to mark it over MSRP at least for a month.  If it follows AMD other product cycles you could expect price to drop to $499 in 6-12 months.

If you have the time you may want to try uninstalling catalyst completely and use cgminer to control the clocks.  cgminer can sometimes (card dependent) get lower stable memclock that Catalyst.  I have found sometimes Catalyst interferes w/ cgminer ADL support.


Weird.  GPU-Z doesn't show the 3 VRM temps?  Maybe GPU-Z needs to be updated to read/parse 79xx bios data.  A side note the reason why GPU-Z has more detailed stats is it doesn't use AMD driver.  It pulls raw data from the card bios and then parses it.  For example VRM temps are available through AMD driver API but they are stored in the bios to handle things like thermal throttling.



Just ignore them.  There is no reason for you to be lying.  What is the angle?  How could you profit from a lie?  The card officially releases in 4 days so I expect to see at least one other member posting about it in the coming week.

Agreed.

http://www.pccomponentes.com/tarjetas_graficas_ati_pci_express/hd_7900/

Simply browsing the site in chrome with translation on led me to this page.

530 euro price is too high. 7970 still sucks at power draw and price ATM. 5XXX still the best for now.

It is (pre) launch day.  What did you expect.  Huge discounts?  The price of a video card never goes up.  They only go down and thus launch day is always the most expensive day for any card's lifecycle.

I think people have gotten unrealistic expectations from deeply discounted end of life 5xxx products.  The 7000 series was never going to be "cheap". 

If AMD was going to sell it at a discount why even release it.  Just save the tens of millions in R&D and just keep selling lower cost 6000 series.  :)

Totally agree on that Death and Taxes,
Since dec 12 - 2011 many shops got some stock for initial sale with the intention to start selling at 9 januari
breaking this rule and amd finds out will be bad for those shop owners caus amd will not quickly give them new supply
With the 9 jan soon to happen the initial prices will be high as allways
But i think they are going to drop soon if amd really gets enough shipped to every shop

I heard from some shops they already got some big piles stocked up, so seems amd got it all kinda done well.
If the initial stocks of the cards sells out quick and amd can keep up with demand, the prices will soon fall to a more acceptable price.

Because sure as hell many hardcore gamers want to get their hands on these very well performing cards

And then the fun starts with the custom build models with better cooling and higher clocks which will this cards make a clear winner
easy overclocks 1100 to 1400 mhz are reported and reviews and some seem to got it at near 1700 mhz

Pics or it did not happen :P

Well said. +1

This noob ain't tricking us with these fake BS results. Next time they claiming a 7970 running at 2000 MHz and getting 800 Mhash/s. LOL

Yeah but how many FPS do you get in BF3?

DiabloD3 has already made mods to his miner for GCN, I am sure he would love to have your feedback on how well it works. See this link (https://github.com/Diablo-D3/DiabloMiner/commit/41dc7645f9cec9fd7a936d7267b48911fdeacffa) for the commit in question.

Tick tock tock ! Let us see some screens.

No worries, AMD's party vans already showed up outside and removed the problem.
http://images.nextnewnetworks.com/83f8b20aa71af28b7728bc7423ddb0c9_blog.jpg

 ;D

Well assuming all the history of comments about the 7xxx are in any way valid then what the OP is posting seems to be quite expected as well.
For the past months I've read that the 7xxx will not be much faster at mining - but it WILL be much lower in power consumption.

1536 Stream Processors would certainly not mean a major difference in the hashing number unless the new Stream Processors were somehow much better.

However the die size drop to 28 nm will mean a power consumption drop.

So can the OP go out NOW and buy a watt meter and tell us what the watt figures are.
Unfortunately you have missed the most important expected feature of the 7xxx series and that is the power consumption drop.

Here's a screenshot of the last few hours' worth of mining.

https://i.imgur.com/nJh4l.jpg (https://i.imgur.com/nJh4l.jpg)

I'm downloading diablo's miner to give it a shot.

F5 F5 F5 F5

Is dat some git log? :D

I guess all the haters can stop hate'in now.   ;D

Downloaded the sources for diablominer and compiled them. ~650MH/s with the default diablominer settings and the card OC'd @ 1125/975MHz:

https://i.imgur.com/f8NnZ.jpg (http://imgur.com/f8NnZ)

~530MH/s at standard clocks:

https://i.imgur.com/DeUxm.jpg (http://imgur.com/DeUxm)

Should be interesting to see the power draw running at 650Mh/s. $/Mh ratio still won't be right for a while but the reduced power draw could make up for it.

LOL. This really is strange. Before people were saying this sucks at mining because of CGN but I see it is quite good performance.

Before people were saying it pulls much more power to hash but now it seems it pulls less. WTF ? LOL. Where is the truth at ?

Fucking awesome.

This is most interesting. That'll be a good run, indeed, when the price matures.

Thanks for the pics  ;) looks good !

Looks like the 5970 has lost it's crown. I'd be surprised if that card pulled more than 200 watts.

Why & Why?

7970
650 MH/s
200W?
$549+

5970
750MH/s
250W
$300 (used)

Now the 7990 will likely dethrone the 5970.  I would guess (based on 7970 results) it may w/ some optimization be the first 1 GH/s card.

As for the wattage.  It is rated for 250W and other benchmarks have put it at 230W under full load.  While it might be less while mining (if mem is downclocked) what you make you think it would be 30W less while also overclocked 20%+?

Guess we won't know for sure until someone gets a Kill-a-watt (hint hint OP :) ).

I thought that because my numbers were wrong. I was thinking this card would pull about 200(because I thought it's TDP was 150) and it's over clocked. My own 5970 pull around 300 to get 750 mh.

WTF and thats with -v 1? 79xx could be glorious if I can find the right settings.

Edit: You know, I just noticed the op post said GPU threads had a viable increase beyond two. Huh. I should go see if that is worth abuse, I never got better performance beyond 2 on DM (its hardwired to 2, and was the first miner to abuse that trick).

The card has been tested under full load pulling 192 watt by a hardware review site who had their hands on the first cards
That full load was doing a heavy game for a certain amount of time, they also mentioned that if you overclock the card it will use more then 200 watt.
And including burnin with furmark now do not tell me you do not know how powerfull furmark is
Its actually more stressing cards then they will ever do in real time usage including mining

This site has allways been in this respect good they test every aspect, and i believe they did more testing then others
So there is no doubt about the results, these guys test with better equipment then a kill-a-watt

Onto the answer with the threads running in the old stucture the alu where not capable of doing more then 2 threads
the new GCN seems to make it possible to make better use of the alu's is stated by amd.
So somehow you can get the card to set every and all alu to work
Which was simply not possible on the vliw architecture
The big increase in numbers on the review (most games) shows clearly that this card easily does up to 33% better then the 6970
I think for the coders out their this card poses a real challenge to get more out of it then with previous cards.
Nevertheless we have to wait what prices it will get at, after the initial run on them because i already know many of the gamers i talked to want 1 or more of them as soon as they can

good! I can't wait to buy one

Curious. What intensity did you run it at on cgminer? I figure the 7970 will be the first GPU to benefit from -I 10. On your screenshot it was only intensity 7!

Mining pulls significantly more performance than games do, and only slightly below Furmark. ~250w with overclock sounds correct for mining.

If sommeone will pull out somme more mhash`s from GCN, it could be best choice.
The prize will cut it a bit untill NV will show its babby.
Theres always hope, that somme gamers will sell somme old and "crappy"HD58XX/5970 ;) to buy HD79XX :) 8)

Thanks for the tip. I just tried -I 10 but it does not seem to have a significant effect on the hashing rate when compared to -I 7.

You should try -I 8 and 9.  It should be about 5% better.

I just tried changing EXECUTION_TOTAL from 2 to 3 on my machine and the hashrate went up slightly from 662.8 MH/s to 666.0 MH/s.

Glorious.

Hah! I thought you might like that hash rate ;)

impressive numbers, when the prices goes down $400 these cards should be hot

Has there been any further improvement/refinement?

now this card is out.. how about some ++real++ benchmarks?
this thing has 500 more sp than the 6990 - how can it possibly be slower????

LOL you don't know a thing about mining, do you ?

6990 has 3 072 or 2*1536

5970 has 2*1600 or 3200 shaders

7970 has 2048 shaders

5870 has 1600 shaders

7990 supposedly has 4096 shaders ?

thanks for repeating what i said.
7970 has 2048 SP
6990 has 1536 SP
thats 500 more. per core, whatever, its more, MORE. so why is the card putting out so much less? maybe if you put your effort into explaining a decent answer (such as 'because the miners need re-optimisation') instead of being a flabberfinger - we could have all benefited.

6990 is dual GPU so has total of 3072 shaders gets about 800 mhash/s using two cores total.
7970 is single GPU so has total of 2048 shaders get about 666 mhash/s using one core total.

Get some sleep dude and stay off SR !

This.

i stand corrected, thought it was also dual.
i feel the need for read

Not that it might matter much at this point but with vectors 4, and I believe 2, to some extent. There is a need to adjust the memory clock in order to optimize it. I am not sure it would even help being CGN. But, if you get time, I'd check it out. Sadly, I've no clue where that thread is at this time. :/

pretty freakin awesome, if you ask me. Now if they can just sell the things for <$400 I'd be happy.  Do you have any TPD numbers with this card?

In the UK currently :

5870 costs 170 GBP and gets 440 mhash/s so about 2.6 mhash/GBP

7970 will cost rougly 430 GBP and get 666 mhash/s so about 1.6 mhash/GBP

Thus, the 5870 is still much better and you can also get a 5970 that gets 850 mhash/s for about 400 GBP.

Power figures ?

Very interesting results! The only missing piece is the power draw from the wall.

My only hesitations at this point are:

1) Price Point/Performance is still super high when compared to used 58xx series cards

2) Lack of optimization in Miners for any new features in GCN/SDK 2.6. Current Miners are heavily optimized for VLIW4/5, so obviously there's going to need to be some re-working for full GCN support.

The only way I can see this card being a viable miner is that it needs to outperform 5970/6990 in performance per watt and $/mhash, otherwise it's just a good excuse to see more 58xx's hitting eBay since gamers will be upgrading...

Thanks for the initial benchmarks though OP! :)

7970 is going to be 200 watts I believe, and the 5870 is 188 (both at stock watts). This is where the 7970 suddenly shines. Even if the 7970 is 250 watts, thats still a jump in efficiency.

You think it will be 200 watts w/ a 20% overclock?  I wish the OP had a kill-a-watt.

Hey OP do you have a kill-a-watt you could purchase locally.  If you are in the states Home Depot and Lowes carry them.  If you can find one locally I am sure we could get together the 3 or 4 BTC to get some accurate power readings.

Thats at stock clocks obviously. I don't know what the mining values will be, all the cards draw less than their full wattage at stock speeds when mining (because large parts of the chip shut off). I imagine 79xx may even get a larger efficiency boost due to this because of AMD's work on power saving, but without a killawatt test, no one knows.

I know that the shaders are used to do the hashing, but is it possible to utilize more of the chip, even if it were at dramatically lower efficiency?

No. I already tried to abuse the texture/memory fetch units, but couldn't figure out a useful way of doing it. Its all fixed function hardware and its not particularly interesting for what we do. Although, I may go try that again, SDK 2.6 seems to be a much better compiler in some areas.

man wat a beast

The kill-a-watt brand doesn't appear to be commercialized here in europe, and I've been searching for an equivalent device locally each time I've had a chance to head out to a store for the past couple of days, but no luck so far.

I also took a stab at modifying DiabloMiner and managed to get it to use 16component vectors, which is what GCN is supposed to be tuned for, but performance isn't what I expect and its really hard to profile/debug the tahiti since I could not find any development tools that specificly support it yet.

Wait wait wait. Are we sure uint16 is such a good idea? Last time I tried >4 (which was before 2.6, btw, I haven't tested with 2.6), it would crash in the compiler. Also, does anyone have a count on the number of registers per CU? There might not be enough registers to handle that.

Also, check some of the larger -vs, -v 40 is two sets of uint4 and -v 44 does three uint4s (unlike cgminer, -v 4 does two uint2s).

I can't wait till the 7990 that is going to be impressive but expensive  :( I might have missed this but what is the heat like hashing overclocked ? and what fan speed

I'm not sure if it's a good idea or not so I wanted to measure it ;) GCN has 64KB worth of registers per CU, and like you said I'm not sure if that's enough. The reason for my curiosity was because GCN's compute units each contain 4 x SIMD units with a width of 16 elements (same size as Larrabee & Intel's MIC, coincidentally), and I recall reading somewhere that each of these SIMD units can retire one 16-way instruction every 4 cycles, so those 16element vectors kind of rang out at me. I also wanted to get familiar with the OpenCL bitcoin mining code and thought it would be a neat exercise (which it was!). Nice code by the way.

I can say for sure that 16element vectors DO compile with the drivers that came with the card.

The -ds code dump for 16 element vectors came out nice and clean, although the last few lines where the result is stored in output seem a bit branchy. It looks something like this:


I tried replacing it with a branch-less expression using shuffle() and vstore16() but haven't managed to get it working. What I've come up with looks something like this:


Anyhow I'm sure that my code modifications are doing all sorts of dumb things. I'm still learning how it all works so please ignore.


I've tried all of the different -v settings available (according to the source) but haven't been able to get any higher than the 666MH/s with the default settings and 3 compute threads.

Well that sucks.  A more universal albeit expensive tool is a clamp meter. 

http://ecx.images-amazon.com/images/I/41F0VY4FHzL._SL500_AA300_.jpg

Is it Mrs. Zhou Tong?

Overclocked @ 1125/975MHz with automatic fan speed I'm getting temperatures hovering 81-83C, and the fan runs at 47-49% speed. You can see some screencaps on one of the earlier pages. But since I prefer lower temperatures and am worried about VRM and memory temps not yet being reported by GPU-Z, I usually run it at 60% fan speed and get temps around 72C. The blower fan at 60% speed is quite loud (its a reference design from Sapphire).

At 100% fan speed, the overclocked card gets below 60C while mining but you can hear it from outside of the house at this point :P, so as lovely as these temps are this is not an option for me as it is also my gaming and work PC.

Yeah, so they still have not fixed that damn reference fan design. Aftermarket coolers FTW !

Damn ATI and their crap loud fan designs :(

BTW, they do make 240v/50hz euro Killawatts, but you might have to order it from the US. They also make 240v/60hz (double hot, like ovens and water heaters) ones and 208v ones for DC shit. Might have to look around, I love mine, its been essential for planning stuff out.

Really nice cards and performance but the price really sucks !

5XXX is much more cost effective ATM. That may change in the future.

Maybe wait for FPGA ?

The branching has ended up becoming the best outcome. It can evaluate those branches in parallel, and you can't easily optimize away branches for memory writes (and theres apparently like 2 or 3 good tricks to get rid of branch waste, its just none of them work on memory writes).

I should look at shuffle. Your way doesn't quite work though, vstore would output H !=0 hashes, which would trigger HW error alerts (and rightfully so) in the host code, and I consider the HW error tracking important. At least, assuming I'm reading that code right, anyways.

I'll go add official 8 and 16 wide support in a bit, should be useful on, say, AVX if you manually enable CPU mining in the code. SDK 2.6's cpu compiler apparently has gotten a lot better from what I've heard.

So does that mean that is the best for 5870 cards ? Or stick to 2.1 or 2.4 ? I am quite confused as to what the best SDK / ati driver combo is ATM.

"Aadamm, what the hell did you DO?, The whole buildings on alert!" "A PANIC ROOM SHES GOD A GODDAMN PANIC ROOM!" "YEA WELL SO DO I ADAM!!"

Notice I said CPU not GPU. CPU mining still sucks altogether. 2.1 is still best for 58xx cards.

That's interesting. Branches always seemed to be published as anathema to well performing kernels. I guess it all depends on how much work is being done inside. For small vector sizes there are few ifs, but with uint16 there are quite a few, so it might be worth investigating there.


Yes, I'm still getting HW alerts and haven't quite worked them out yet. I posted the snippet earlier from memory and missed a couple of steps. The latest (broken) code I'm working with looks like this:


That "if" might be totally unneccesary, and I still don't quite understand how the output array works, but it might give you a better idea of what I was trying to do to avoid all those branches.


I'll be watching the repository then :) It should almost definitely help with more modern CPUs and Larrabee/Intel MIC.

The output array is basically a massive hack to prevent multiple outputs from hitting each other, although the chances of getting multiple outputs is extremely low. The size of the array now is massive overkill, but it also seems to be a strangely optimum size for hardware.

Now, what would give me the most benefit is some way of sorting the outputs in a single cycle so that the pair of { nonce, H } could instantly give me the best nonce, and then only evaluate that. There seems to be no way to do this (and yes, I imply reverting that one bit of math so that H == 0 is literally done at the end again, makes it much easier to sort on shit). The nonces themselves can't be sorted because its completely random, they're meaningless values essentially.

So what do you think we can get out of this card being 100% optimistic? How much is it limited by the current best software solution.

666+mh/s
some hardcore guys used liquid nitrogen cooling and overclocked it by 84%

Up to 670MH/s @ 1125/975Mhz now with the driver AMD published yesterday! And I finally managed to find a wattmeter, so you can expect some measurements later today when I get back home from work :)

Yay!!

I've measured my system and these are the results:

	Stock (925/1375MHz)	Overclocked (1125/975MHz)
Mining                        :	371 W @ 550MH/s	385 W @ 670MH/s
Idle                            :	118 W	118 W
Difference_(gfx_card_W):	253 W	267 W
MH/J_(system)             :	1.48	1.74
MH/J_(gfx_card_only)    :	2.17	2.51
MH/$_(gfx_card_only)   :	1.00	1.22

(MH/$ estimated using lowest listed price for HD 7970 on amazon.com today)

Excellent findings!  Thank you for your candor!

-EP

Thanks for putting that all together.  For lack of a better term.... brutally bad.

Significant reduction in MH/W compared to 5000 series. 

anyone care to update https://en.bitcoin.it/wiki/Mining_hardware_comparison  with the new findings??

-EP

Not entirely. Remember, this card will need significant optimizations, and don't apple/oranges vs 58xx if you're not using the same SDK. Nothing is going to beat 58xx on SDK 2.1, and you shouldn't expect anything that glorious ever again. Its a classic. That said, SDK 2.5 on 58xx, you lose about 4-5% give or take, dunno about 2.6, still haven't quite figured out how to best fix that.

Could be worse, I have a GTX460 that makes me want to tear my hair out :-\

Thanks for the numbers, I've got one on the way.

As a reminder, 100% fan speed is a good way to kill the fan, they were never meant to be ran that high.

Don't go above 85%.

I've downloaded Sapphire's TriXX software and pushed the clocks even further, even managing to get the memory underclocked down to 150MHz:

	1150/150/1.17V	1175/150/1.17V	1200/150/1.17V	1225/150/1.175V	1250/150/1.2V
Idle                   :	118 W	118 W	118 W	118 W	118 W
Mining                 :	392 W	400 W	408 W	415 W	441 W
Difference_(gfx_card_W):	274 W	282 W	290 W	297 W	323 W
MH/s                   :	675MH/s	690MH/s	705MH/s	716MH/s	733MH/s
MH/J_(system)          :	1.72	1.73	1.73	1.73	1.66
MH/J_(gfx_card_only)   :	2.46	2.45	2.43	2.41	2.27
MH/$_(gfx_card_only)   :	1.23	1.25	1.28	1.30	1.33

Power draw reaches a tipping point around 1250MHz where the core voltage needs to start getting tweaked quite a bit for stable overclocks. I have a feeling that with more exotic cooling or insane fan speeds (I stuck with 60% which as I stated earlier is already quite obnoxious) that this card is sure to go much higher. The tool didn't let me lower the voltage either, which might be an interesting thing to do to increase efficiency.

1125Mhz still has the best efficiency with 2.51 MH/J.

For the price I would rather get a 5870.  I'm able to get 440 out of one of them and its more then half the price. 

I have a trouble with Radeon HD 5850 and 11.12 driver (with integrated AMD APP SDK 2.6). Linux. It is 360 MHash, but 11.10 + AMD APP SDK 2.5 gives 412 MHash. It can be a reason of slow speed of 7xxx.

We grabbed a 7970 today to play with.. still on the fence if I am just going to sell it or not.. but I can confirm that the numbers in this thread are in line with what we are getting..  I have only pushed it up to 1125mhz and that's about as far as I will take it and no overvolting, but I can confirm it seems to run quite nice there doing 670MH all day..

teek

Doesn't seem worth it with those numbers, maybe an undervolt could improve the mhash/w.

just wait until there's optimized miner & drivers.

Of course the initial price isn't really a major issue unless it is way above current hardware
It's the long term running cost (unless you don't expect to run it for very long)
Even these FPGA's are priced well above the ATI card prices for similar hash rates, but since the FPGA's use a fraction of the power, long term is the reasoning behind buying them.

Wow, that's pretty impressive!  733 MH/s is really close to what my 5970 puts down, albeit the 7970 is drawing over 100W more power at that hashing rate.  Still though, that's really good for a single GPU. I can't wait to see what a 7990 will do!

5970 does 300w stock on gaming, 7970 does 250w stock on gaming.  At ~300 watts, the 7970@1225mhz is doing 716 mhash, the 5970@725 on SDK 2.5 is doing 646, 671 on magical SDK 2.1. The 5970 has the advantage of being undervolted over the rest of the 58xx family and being able to run SDK 2.1.

Given that, I think AMD has produced a very impressive chip.

I'm not sure where you got that ”300W on gaming figure”, but I'm just going by my killowatt meter. At idle my system draws about 140W. Full out mining (810/200/1.050V) @ 740 MH/s it draws 350W, which tells me (using 1onevvolfs math of mining-idle=card wattage) my 5970 is pulling about 210W -- about 100W less than the 7970.

Worth noting:

http://www.anandtech.com/show/5261/amd-radeon-hd-7970-review/14

AFAIK, 5970 takes close to 50 W in idle, and 7970 takes 3 W.
7970@1225/150/1.175 = 415-118+3 = 300W@720 mhs
5970@810/200/1.050V = 350-140+50 = 260W@740 mhs

294 and 250 are the official AMD quoted figures for maximum draw on 5970 and 7970. Mining uses less power due to parts of the chip shutting off (texture units, etc), and I suspect GCN has superior power savings over 58xx in that area, if not, they're very similar (ie, 5970 doesn't draw 294 while mining at stock speeds, and 7970 doesn't use 250 by the same amount give or take).

Hello all, Got my 7970 but I'm a total noob to mining. Kiv's GUI miner won't work with the 7970 and I'd love to try Diablo3D's miner but can't figure it out, anyone willing to help me and I'll throw a few BTC their way when I can mine a few. Thanks.

Chapter 4.16 of the AMD OpenCL programming guide (http://developer.amd.com/sdks/AMDAPPSDK/assets/AMD_Accelerated_Parallel_Processing_OpenCL_Programming_Guide.pdf) has some optimization guidelines for GCN. This snippet explains why my vectorization experiment didn't help performance:


Oh well, back to the drawing board!  ;D

Hi, would you be able to help me set up my miner?

What's in it for us ?

On another note, 7970 and I believe the rest of the failure that is CGN will be crap for mining as far as I can see.

More power usage than 5xxx series -> crap MH/s to power consumed ratio
Much more expensive than 5xxx series -> crap MH/s to price ratio

So much for AMD's "Fold and mine faster than ever with AMD App Acceleration powered by the unprecedented 28nm GCN Architecture."

I am waiting for Nvidia to finally step up their game and come to the mining table because they are totally missing out. Their integer performance just sucks !

Also, I am quite happy because my stock of 5870s is just getting more and more demand and valuation by the day with these CGN disappointments.

Roll on AMD, to your grave !

Dare I state the obvious in reply to above ... the 7970 is a single GPU card ...

There is nothing false about their statement.  It looks like the 7970 IS the fasted mining card.  AMD made no claims in terms of efficiency (MH/W or MH/$).

I mean it is like buying the fastest sports car and then saying the company lied because it costs more and has worse gas mileage than a Honda Civic. :)

I've just found a forum post where someone reported stable performance with 0.939V core voltage at stock speeds (http://www.computerbase.de/forum/showthread.php?t=1008559). Stock voltage is 1.170V so if we ignore the other components of the card that would mean a ~20% power reduction (or more, due to reduced temps), and therefore a 20% efficiency gain as well (bringing the MH/J up from the 2.2-2.5 range to the 2.75-3.125 range). They also report a 7C temp drop which should let me reduce the fan speed from an annoying 60% to sub-50%, which is barely audible against the other case fans.

But this is all just theory until someone actually measures it, so off to find a BIOS that actually lets me lower the core voltage!

Actually power decreases at the square of voltage change.  A 20% undervolt is huge.  If true one would expect more like a 35% drop in power.  Now let me hedge that by saying dropping core voltage doesn't drop mem voltage so it is more like a 35% drop in CORE ONLY power.  Still overall power savings should be >20% for a 20% undervolt. 

If true that could make an undervolted 7970 (especially if someone can figure out how to undervolt the memory something I haven't been able to do on 5970) a hybrid between FPGA and GPU.

Nail, meet head.

I'm not going to believe that until I can prove it. Also, mining kernels are very abnormal for how kernels function, it is very much not a typical kernel. I'm glad they updated the programming guide for GCN though, its very useful for intermediate OpenCL programmers.

This. 7970 is the fastest GPGPU bar the 7990 coming out RSN. AMD hasn't misadvertised anything.

Power usage in Mhz is linear (ie, 500 to 1000 mhz will double power usage), but power usage in voltage is squared (using 100 watt at 1 volt? It will use about 150 watts at 1.25 volt).

I just pushed a new commit. You can do vector layouts instead of the old bitmasks. So instead of -v 18, you do -v 1,1, instead of -v 36, you do -v 4, instead of -v 40, you do -v 4,4, instead of -v 4, you do -v 2,2, etc.

Yes and he indicated he UNDERVOLTED by 20%.  0.8%^2 = ~0.65 or 35% reduction in core power at the same clock (assuming the same clock is possible w/ 20% undervolt) which is why I was impressed.  A 20% undervolt is kinda hard to believe.

Yeah, but you traded top speed for efficiency, which lowers efficiency in the dollar sense (ie, mhash/$).

Managed to undervolt without flashing the BIOS. Just needed to enable the core voltage option in MSI's Afterburner software and that's it!

At the moment I'm running overclocked getting 670MH/s @ 1125/975 AND undervolted from 1175mV to 1112mV, for a 37W reduction from stock settings down to 230W power draw for the card at this speed. The fan is also down to a more bearable 50% ;)

So with these stable settings the efficiency figure is 2.91 MH/J. This is smack in the middle of my estimated figures from earlier. I'm sure that lower core frequencies will allow for even bigger undervoltages.

The funny thing is that the card would actually run stable at 1040mV with a 197W draw as long as the only thing it was doing was mining which is 3.4MH/J. But fire up a game and it would crash after a short while. Maybe this is due to the power management features turning things on and off as was mentioned earlier in this thread? Anyway, it seems that head-less mining boxes using these cards have room for more efficiency gains.

Tomorrow I'll measure some more data points to get a better idea of what these characteristic curves look like.

P.S. I dunno if its a bargain for mining or not but I'm having a lot of fun with this card so far and mining quite a few bitcoins in the process :)

Thanks for providing these numbers for those who are still waiting on their cards, or are considering picking some up.

Not bad for an AMD PR flack! ;)

:o wow, nice! It working with a bitcoin miner, and crashing with a game is certainly because the bitcoin miner ignores large portions of the die, causing them to use little or no power since they are idle. This is why it may have been stable at that dramatic of an underclock.

That's amazing, considering how much better this card is with games alone, sounds like a very good all around card. I'm sure the prices will drop significantly when the fab can spit out more cards.

It's mostly due to the games utilizing the card's ram. Which of course needs more juice.

Awesome results though thus far. I'll have to go back and poke at my lil 5830's a bit but I don't believe they undervolt nearly as well as what these 7970's are doing thus far!  That in itself is promising.

Thanks for sharing your data with us, m8.

cheers

Got my Diamond 7970 in today.

Running latest DiabloD3 (https://bitcointalk.org/index.php?topic=1721.0), I'm getting ~593Mhash/s @ 1000mhz with default settings. Have not started tweaking yet.

Fan running at 70% has the card at ~71C

I'm running my Sapphire 7970@1125 with temps at 74 with 55% fan.

 :P ;)

getting 668Mh/s

I've got about $6,000 at my disposal and I'm trying to decide if I want to buy 8 more 7970's or buy the bitcoins outright.
Anyone here willing to sell me 1,000 bitcoins for 6 grand?

It's probably the location of my room (stale air), but I'm getting 87C temps at stock (1000Mhz black edition). :(
 
I even have it in a large Corsair 800D which has pretty good air flow...
 
I'm also contemplating liquid cooling (for other reasons) but there are water blocks out for the 7970 so that would help if I decide to go that route.

That is a no brainer.  Buy 20 used 5970 and get 40 GPUs.

 B-b-bu-but 7970s :*(

I'm having a hard time finding a good source for 5970's.

I was under the impression from reading in these forums that an under-volted 7970 is at least on par with a 5970 or a little bit better.

But wouldn't just buying them be easier?
.

Buy FPGA's Some of us gamers still want those cards for gaming xD
Or buy some used 5970's

Yeah I'm thinking about that as well for my next upgrade. With a waterblock and the right bracket these cards fit in a single slot and overclock so much better. And best of all no more annoying fan!

Woah, and I said its probably possible for >600 on stock clocks. So close.

Can people post room temperatures also with gpu and %fan. As room temp has pretty huge impact on %fan.

Yeah i agree,
My temps waver by an entire 5° throughout the day, And somepeople may live in cold climated where there "normal" temp inside thier how's is 18°
I live in canada, My home's temp is average 19-20°, Feels hot.

Sapphire 7970 - Using phoenix 1.7.3
I'm getting:
550MH/s on 925MHz GPU / 340 MHz memory
712MH/s on 1200MHz GPU / 340MHz memory

Bumping mem to default 1375MHz adds like 5MH/s so I don't really see the point.
I haven't found the lowest stable point for GPU voltage yet, but it seems to be approx 1000mV for 925MHz/340MHz.

Now waiting for optimized kernels :)

My Catalyst GUI shows (in the tachometer-style pic) that my memory is way up to default level, even tho I've decreased it down to the lowest value.
Is this normal?  I'm wondering what I should do to correct it if it isn't.  I think I'm on the newest drivers (12.1?)

The Catalyst GUI will only respect underclocks to 975MHz AFAIK anything lower will get you default memory clocks.

ugh, why do they even give you the option then?  That's so <insert insulting remark of your choice here>.

Inserted

712  MH/s is better than the 5970 in mh/w I believe, and nearly, if not beats the 5970 at stock.

-EP

Lolz.  As if you can even sniff a 5970 for $300. 

BUY THE BTC!!!  Unless you're getting free power, the ROI for $6,000 worth of graphics cards will take so much longer to re-coup vs. any value that BTC rises = profit.

-EP

Everyday on ebay.  Strangely prices have gone up in the last month.  Looks like it is more like $340-$360.  You would think w/ 5970 supply > demand and prices would be falling.

Sure at 60% to 80% higher cost.  5970 can be underclocked also. 

Every 5970 I have seen in the last 2 weeks is 350+ on Ebay.  I know, I have been bidding. 

driver 12.x needs a hack in the register to make the memory and proper downclocks possible
google for it you only have to edit one line in register to make it work

Nah,
$500
http://www.ebay.ca/itm/ATI-5970-2Gb-512Bit-DDR5-DirectX-11-New-warranty-1-year-/270885811822?pt=PCC_Video_TV_Cards&hash=item3f120d766e
New.

Any hints on where to find this info?  I must be using the wrong Google terms because I can't find it.

You cannot find it because what he says is not true, I've been running with 12.1 eversince it came out, Nothing changed as far as OverClocking is concerned.
Just use MSIa, It has a backwards compatable "bug" in thier "unofficial overclocking" that allows you to Attempt to set Any clock to Anything

win7x64 Cat 12.1 SDK 2.6
crossfire6870 990core 490mem <---Sweet spot for mining
Hell, for awhile i had the cards running at 1020 core and 300mem because i was told 300mem was the best, It's the Most Common best, It's quite specific for the cards.

CGminer also changes clocks Flawlessly
I also hear (NEVER TOUCHED) that Shappire Trix is a good tool.

Doesn't seem to work for the 7970, though.  The slider won't go below 685 memory clock.

Hmmm.
I shall just assume i do not know what i am talking about since i do not own a 7970

well, right now they are all reference except for one XFX card.

Just wanted to chime in on this. I just picked up a pair of Sapphire Radeon 7970 GPUs. I've overclocked them to 1125MHz. I'm using Diablo Miner for this run. My hash rate ranges between 1335 and 1340 MHash/s. Note that this is with dual GPU.

EDIT: I'll try and get power figures soon. I have a Kill-a-watt, I just gotta find it.

EDIT 2:

System Specifications:

Intel Core I7 960 @ 3.2 GHz
24 GB DDR3-1600
2 x Sapphire Radeon 7970
Corsair 750W PSU
WD 2TB Black HDD

Power Consumption in Watts w/ (MHash/s):

Idle: ~130W (0 MHash/s)
2 x 925 MHz: ~670W to ~680W (1104 MHash/s) - Power use increases as temps increase
2 x 1125 MHz: ~750W to ~790W (1345 MHash/s) - Power use increases as temps increase
Idle (with OC): ~130W (0 MHash/s)

I'm guessing that a slightly higher rated power supply with better efficiency might be called for on this one.

No shit. Corsairs are rated at continuous wattage, and 7970s use 250w each at stock, and you shouldn't exceed 90% of that 750, and you're doing 750-790 at the wall? Holy crap dude.

Go buy one of these: http://ur1.ca/7sqrz

Weird short-URL. Redirects to:

Yes, I used a tinyurl because the newegg url is so long.

Keep in mind that the draw from the wall will be about 20% or more greater than the DC load on the PSU. That 750W number is what it is capable of delivering in DC. At 750W at the wall, the PSU may only be delivering around 600W to the hardware (more if the efficiency is better).

Around 675, iirc their peak efficiency is above 90%. So ~610 watt at the meter is what I'd consider safe.

You could probably shave another 40-60W per card by undervolting. I got 50W off of mine @ 1125MHz which lowered the temps enough (8-9ºC IIRC) to afford running the fans at slower speeds, and others on the forums have been able to undervolt even more than that.

D3, I think you might have your numbers backwards.  Power at the wall is always higher than the power supplied.  Power supplied is what the PSU is rated for.

Plus, peak efficiency is typically around 50-60% load.  And I don't think there's any 750W Corsair PSU with that kind of efficiency.

Since he didn't actually say what model it is, I'm going to assume a TX750 V2.

Quality review here:
http://www.jonnyguru.com/modules.php?name=NDReviews&op=Story&reid=230

So at the power draw he's at, he's MAYBE getting 85% efficiency.  So 790W draw at the wall is about 670W supplied by the power supply, or a little less than 90% of the power supply's maximum capability.  It's not the most optimal situation, but he's well within the capability of the PSU.  If it was a lesser brand, I'd be concerned, but I wouldn't be too worried about the Corsair unit.

Yes, but what comes out of the wall is what is being used total. So. Something.

Did I mention I hate math?

I found that the lower the GPU clock is, the lower you can take down your memory clock. Works the same the other direction. Currently I have mine at 1100MHz GPU and 950MHz. That's the lowest I could get my memory clock with the GPU clock like it is, without it jumping up to stock speed. Fan is at 61% and temp is a solid 66C. Average 650 MHash/s. I wish I could figure out a way to lower the memory clock more to keep temps even lower so I can turn the fan down. I can't undervolt either because even in afterburner, after I enable volt changing, it won't let me change the voltage. Any ideas on that one?

Hmm. Might try a 7970 out! (Me and some guys are building a rig).

Out of curiosity, does anyone know if I can use a garbage CPU/mobo/RAM? (Ya, I'm a noob at mining).

Yes, CPU plays only a very small part in mining on linux ...
CPU's will typically downclock while mining and still used < 5%

Yes my rigs use Sempron single core (watch out the OEM version has no fan), 2GB cheapo ram (1GB might work but 2GB is so cheap), no HDD, and 4GB USB drive running Linux.  It was Linuxcoin but now is the Linux install kano has in his signature above.

Thanks! I was actually thinking of using a Sempron! Thanks for the help!

1225MHz @1225mV / 366 MHz (1500mV)
seems to be high as I can go without display artifacts.
This yields avg ~725MH/s and temperature is 80 degrees on fan 55% setting.

Phoenix settings:
-k poclbm WORKSIZE=64 VECTORS1 DEVICE=0 AGGRESSION=11

nice man! good luck

Anyone know how to get the memory clock on the 7970 lower? If I lower it more than 950 mhz, it jumps up to stock speeds. I can't control the voltage in afterburner, and Trixx won't let me lower it, only raise it.

Have you enabled UnOffical Overclocking in MSIa?

I enabled voltage control but it still won't let me change it even after restarting the program. I didn't see the option you're talking about though.

http://forums.guru3d.com/showthread.php?t=338906
Here, You need to edit the config file

Thank you

Damn noisy at 55% for anyone who is close to it. :)

True, I'm using 925MHz 1020mV / 366MHz 1500mV and fan @29% whenever I want some peace and quiet :)

Has anyone found any optimized kernels yet? Or manage to get more than 550MH/s @ 925MHz ?

Version 2.2.1 of cgminer controls the memory clock speeds just fine: I run my clocks at 150, something Catalyst Control OD confirms. I have found the best results with my two 7970s on cgminer, getting 449 (intensity 6) and 480 (intensity 14) on the two, respectively.

And they're great for gaming. Viva la ATI!

Those are great results for that card. Doesn't surprise me a bit. These cards are utter monsters. I'm also curious about the power draw.

Hmm... the best result is 550 at least on stock clocks.

I really hope they come with the 7990 as they did with the 6990, if they can crank out 2 x 7970 performance we are talking about a graphiccard which is usefull

I have a 2 x 7970 set up, 1049mV, 1105 core, 685 clock. Draw from the wall peaks at 580watt on load.  Just a reference.

I have not tried yet to lower the power consumption on my card, but I have it overclocked to 1275 MHz core, RAM is more difficult to modify with software, but using the new bitminter beta just released I can achieve 780 MH/s; GPU-Z sensors say my overall power consumption is 206.3 watts; so if my calculations are correct this is 3.78 MH/watt. This card is can far more, with the new AMD drivers I can pull 833 MH/s at 1310 MHz, but it will crash in a few seconds unless I raise the voltage.

Just wanted to throw my .02 cents in here.

I now have 7 of these bad boys up and running, and I think I have found a sweet spot for those running them enclosed in a case:

1150 core
600 mem
Cgminer
aggression @ 10
Fan @ 70%

688 MH/s, 75 degrees, rock solid stability. 


I would love to throw a waterblock on one of these things and see how hard you can push them.  I really don't think 800 Mh/s is out of the question. 

Nice. What kind of watts like this? 

70% fan is too high for my preference.

My sweetspot is

1125/685

666(lol)mhash

55% fan

68C

It would be really sweet to see a single 7990 around 1.4 GHash! Can't wait to see the results once they come out.

I doubt it.  Much like you don't see any 5970 @ 840 MH/s.  They are same chips but they are clocked lower and worse they tend to have inferior VRMs.  5970/6990 use 3 phase power,  5870/6970 use 4 phase power.

Maybe 1.1 GH/s.

You could easily drop fanspeed to around 40-45% or increase voltage and GPU speed.

These GPUs can handle even 80°C. I take it that you are using only CCC to adjust clock speeds.
I recommend using MSI Afterburner or similar tool.

840 MH/s out of a 5970 *can* be done.  I've done it.  However, I didn't like what I saw on my Kill-a-Watt though since it required overvolting.  Technically it's possible but then again so is snorting wasabi.

my sapphire 7970 3gb oc does not really like memory lowered
putiing it on faster speeds is no problem factory wise it runs 950/1000 on the core and 5700/5800 (1425/1450) on the memory
somehow this card loves higher memory speeds because it goes easy to 6000+ (1500+)
but since that produces a lot of heat i stopped at 6060 mhz
anyway when i lower the speed on the memory under 1200 mhz my pc starts to crash when i run games
i can set it at 1000 for instance and will run but programs simply crash on different moments
upping the mem and all runs stable as a rock

currently i run at 1100 core with mem at 1200 producing 653 Mh/s
 

I though increasing the memory was counterproductive in power consumption, despite the latest SDK on 7970 having a bit of dependency on mem speed. Though not very counter productive, but a few watts get reduced on a lower mem speed.

Yes you asumed correct for me and my high electricity costs its allways been the goal to undervolt / underclock at best hash result.
But i also want to keep gaming and yes i want to game when mining for that i reduce the graphic needs in the games to about medium settings, which does not affect my mining at all

So yes i can play World of Tanks at medium settings and the miner does not even show a small bump 

However this far i am very pleased with the new driver set from Amd.
Installed catalyst 12.4 and guess what it runs stable now even at 800-1250 on the core and 800-1550 on memory (min -  max)

To make matters clear as soon as i set the memory below 800 the crashes appear again, that is when mining and using some programs like games, video processing tools and sometimes even with winzip or rar

My cards also tends to crash if the difference between core and memory speed becomes more then 150 mhz again speeding it up does work fine but when trying to lower the speeds the crashes start.

However need todo some more tests with the newer driver and cgminer 2.3.6 to see if i can now go below the previous limits.

This far 800/800 has been the most silent, efficient and coolest miner setting but time will tell (74 c gpu temp).

looks like the diablo miner is the  best miner to go with

bronan, I can confirm your memory speed theories.  My 7970 is unstable if I set the memory slower than stock.  My rig crashed 5 times a day until I experimented with putting the memory at the stock speed.  The rule of -150MHz from core clock doesn't apply to this card.  Wish it was like my 5830's where I can set the memclocks at 300.

Very cool. Im waiting on 2 cards in the mail so i can start mining aswell

Haha yes both my sapphire 5870 vapor x went to the usa but where perfect able to run at 0.900 volt 750 mhz on core and 166 mhz on the memory
They are still awesome cards and i think the new owner is very happy with them.
I remember i often bend over towards my computer to make sure they where running fine :D ( so silent )
And yes i could game on them also while mining with no problems ofcourse not with the graphics power of these but still.

Do not get me wrong the 7970 i have now are monsters and i love them but underclock/volt wise they are a pain in the ass they just want speed and more speed :D
So a bit underclocking is possible but not much, undervolting does not work either, so far it seems that the cause of all this is ELPS the advanced power save build into the cards. This far i have not been able to turn it off completely i would like that though so i can proof if my theory is right.

Maybe someone has advanced knowledge at ati bios matter and would like to help find out if its possible to turn it of by flash the bios with a alternative one

When will see new ati drivers? I made 700-750Mh with my Sapphire 7970,  but I think it can do much more with better drivers, or maybe new version of Diablominer.

PLEASE share how you get 750 because nobody seems to be able to get that much normally !

Thanks !

well, clocked near 1300 core. Using diablominer

I have never got more then 704mhash out of any 7970.

Doubt it is stock clocks.

Card will soon die unless you got proper hardcore cooling :-\

What was the shares/minute over a 24 hour run?

They all basically produce 0.6 MH/s per Mhz engine clock speed.

Nice, so at 1250 MHz you should get 750 MHash/s ?

Too bad the ASIC is coming :D

That seems about right. On DiabloMiner on my 7970 at 1125 I get 675.

The bigger question now is : how many reference 7970s ( or the new better binned GHz editions ) can reach 1250 MHz without overvolt ?

I dont get this binned GHZ edition shit. Mines a ref design factory overclocked to 1010mhz, and I got it like 4 months ago. It happy cranks away at 1125mhz at stock volts.

You don't get marketing ?

:D

GHz edition is supposed to make it "cool" to the noob gamers and convince them that 7970 is comparable to GTX 580.

When it clearly isn't ...

Of course it isnt. NVidia totally botched the 580, they'll never catch up to AMD at this rate.

So, why Nvidia are not getin' into BTC at all?

Because it would be a stupid idea, AMD isnt even "getting into btc"