Bitcoin Forum

Previously I had a 5830 running at 272 Mhs with Phoenix 1.7 Phatk2 128worksize on Cat 11.11

Now getting 250 Mhs with the same settings.


Ouch.


Anyone esle try the preview? Going to try installing the old Opencl Package from 11.11 ontop of 12.1



(I was able to bring it back up to 265 Mhs with the original Phatk kernel, but still not as nice as 272....)

same speeds for me :-\

With both 11.12 and 12.1 I went from 310Mh to 270Mh on 3x 5380's clocked at 950/300, AMD x4 w/ Win7 64x.

I'm using 2x 5850, but i don't think there was a hashrate drop. MAYBE 10 mh/s drop, max.

my 5850 speed dropped 20 mh/s. im trying using driver sweeper and install previous driver still had the same problem  :-\

from 11.11 to 11.12

plus:
first time multi-gpu-cpu-bug solved in 11.12 - for me
installed on 2 x 6990-rig

but

minus:
hashrate dropped ~5 %

12.1 is only preview and seems to have the performance of 11.12

When I installed 11.12, I noticed the opencl version was hugely different. it went from 2.5.793.1 to 10.0.831.4. opencl is probably a completely different beast now.

edit: huge mhash drop. asus 5870 / sapphire 5870 / sapphire 5830 xtreme went from 475/501.8/317.1 to 440.1/454.5/286. DO NOT INSTALL!!!!!!!!!
edit2: maybe you can update to 11.12 driver but stay on "old" opencl. i'll post back with results.
another: I didnt even check to see if cpu bug was still there with 3 vid cards. sorry guys
zzz: after uninstalling, rebooting, driver sweeper all amd products, rebooting again, installing 11.11, I still have the "new" opencl somehow. my good miner has been tainted!

Yeah, I noticed this aswell, I was actually JUST about to update all my drivers on my gaming rigs to play battlefield 3, Hearing that "12.1 will probly finally use all of your GPU in BF3"
But seeing this performance drop aint worth the headache...

Anyone got info on that part?
Running crossfire 6870's 1ghz core each, 11.11b (seems the best as far as i know) and battlefield 3 with:
2xAA
1920x1080
Shadow=High
Texture=High
Postprocss=Off
Motion blur=Off
Everything else, Ultra.

And yet i get a Steady Solid 45fps, That only drops to <30 when i fly and look over the Whole Map
It feels Correct, There doesnt FEEL like and performance is going to waste, it FEELS like it's just NOT being asked to get used
Like fucking shit i can mine with -f80 and my fps doesnt drop too much
And i'll get about 80mhash/sec percard!!!!

With 58% load on both cards.......................................................
I WANT TO TURN MY SETTINGS UP THANK YOU VERY MUCH!

oh my god I have tried like 4 times now to revert to 2.5 sdk and i cant revert! I've tried uninstalling -everything- amd related and used driver sweeper every time too, but i simply cant revert it! its like it just decides to keep the most recent sdk "just in case".

i'll pay bitcoins to someone who can help revert without needing a complete system format.

Dual boot linux ... ok how many bitcoins? :P

/nope. :P not touching linux.

Anyways it felt like forever, but I was able to revert to 2.5 SDK. Bitcoin offer is no longer valid. If you're using windows, simply replace these files with those of 2.5 SDK and you should have your faster mhash back

C:\Windows\System32\amdocl64.dll
C:\Windows\System32\amdoclcl64.dll
C:\Windows\System32\OpenCL.dll
C:\Windows\SysWOW64\amdocl.dll
C:\Windows\SysWOW64\amdoclcl.dll
C:\Windows\SysWOW64\OpenCL.dll

I also uploaded a rar file with these dlls if anyone needs them to revert. http://dl.dropbox.com/u/9768004/AMD-APP-SDK-v2.5.rar
just extract it to the drive you have your OS installed on and it'll extract each file to the correct path.
edit: forgot link.

thx for this detailed infos...

the "old" performance is back again  ;D

it is slower with phoenix but it has small boost with cgminer with flag -I 9(to remove cpu bug)

w7 x64 5870@950/300 gives 434,4Mhash/s, with catalyst 11.11/b/c/ was 433,3

One mega hash from One user...
Anyone else got something on the positive side for these drivers?

It's very likely that nobody will see improvements until kernels get rewritten to better accommodate the new version of opencl and AMD APP.

\


Same here...  and it didnt solve the cpu bug for me.

I already figured out how to revert back to the faster SDK. Just scroll down. I even posted a compressed archive of the files you'll need to revert back to 2.5 (2.6 is the new, slower, sdk)

Maybe for mining coins, But perhaps the new OpenCL runs games better

There are extremely few, if any, modern games that utilize opencl. OpenCL is aimed more towards scientific applications like WPA cracking. If a game needs numbers calculated for something like physics, it'll either be offloaded to the CPU in order to keep game framerate high, or if you have an nvidia card and the game specificly uses physx, it'll be offloaded to that instead.

You're probably thinking of OpenGL, which coincidentally got a version bump to 4.2 in 11.12 and 12.1 preview.

Nah, Definitly know the differance between openComptuingLanguage and OpenGraphicsLanguage,
Why would you think that it would be offloaded to the CPU unless you have an nvidia card running PhysX?
In BF3, My cpu is up around 80% usage on each core (PhenomIIx6@3.6ghz) While my dual 6870's are just kinda "lulling" at 60% while they run the graphics for battlefield.
I would assume that OpenCL would take some of this CPU load off, And run it in my GPU.

i also have a slight drop in hashing power since the new drivers but the CPU bug is gone

The CPU bug is gone?!?!?!! Really?! Your not bullshitting?

So guys, take a look here, Apparently AMD notcied the Failure of 12.1...
They've now got some BIG RED TEXT at the top
However i still fear TOUCHING my BF3 drivers.... Last fucking thing i want is Stuttering


Info on the BIG RED TEXT? Was it there before? Is this a non-mentioned hotfix/quickswap?

Yup, CPU bug is finally gone, i'll take the small drop in hashing power for that :) I play BF3 and I haven't noticed any drop in performance in that game. However the previous driver version gave me horrible dissapearing of objects/terrain in skyrim that was fixable by minimizing the game then maximizing it- got very very annoying.... that has since been fixed with my update yesterday

I should mention that I am using the 11.12 drivers that were recently released

Thanks, Updating my sh!t now, I'll let everyone know.

How many cards? Also are you using really high intensity/aggression, or somewhere in the middle/low end?

Two here, CPU bug Removed? YES 100% removed for me, YAAAAAY NO MORE 100% ON ONE OF MY CORES!

Running dual 6870's, Lost 10mhash/s on each card. Sad...... But worth it.

BF3 runs the same

HAAAAKUUUUNAAA MUUUUTAAAATAAAAAA!!!!!
IT MEANS NOOOO WORRRIIIIIIIESSSSSSS FOOOOORRRRRRRRRR THE RESSSSSSSST OFFF YOURRRR DAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAAYYYYZZZZZ

Ah crap, Now i have a worry,
Using these drivers while mining at -f120, Using the internet for video/etc was all smooth and dandy, Now it's kinda....
Stuttery........
Help?

I'm using 2 cards- a 5870 and a 5770- i use -f80 when I'm using my rig but otherwise i just use the standard -v -w64 settings when i'm away or sleeping.... i'm just glad that cpu bug is gone- it was like i only had a tri core while mining the last half a year :D

Has anyone written a gui for CG miner? because using GUIminer(poclbm kernal) my hash rate drops by an annoying amount

I used both CGminer and GUIMiner and I got the same drop with each. I think it's card dependent.

My mhash dropped on all my boxes with 11.12/12.1. Also high aggression/intensity cpu bug still there.

Updated to 11.12 except for the app sdk. GPU usage is higher than before. I had to decrease intensity to keep the cards at the same temperatures they were running before the update. Speed is the same as before, CPU bug is gone.

It would be prudent to post which series and how much of a perfomance drop you experience, on 5770, it was good for -15%. Perhaps 6xxx is more optimized (or less unoptimized) in this OpenCL.

Please see my first post on this thread, I posted that already.  ;)

Since my first test I've also installed 12.1 on another machine with 3x 6870's (everything else is basically the same). I got a similar performance drop.

It's interesting that the performance is coincidentally very close to what each card would do at stock clocks. I was suspicious that maybe something with the new driver was interfering with the over/underclocking. But GPU-z reports that the cards are being clocked correctly after I dance my many-step OC/UC dance.

I'm now running 11.12 with SDK 2.4 and I'm experiencing lowered CPU use (only 60% now while mining) while retaining full performance from 11.11.  I'm going to try going back to SDK 2.1 tonight when I get home and see if that will get me back to <5% CPU usage.

im running 12.1p with 2.4 sdk and i get 0-6% cpu usage now and it stays closer to 0 most of the time pretty sweet i did lose some perf but not much to even worry about maybe 5mh.

http://i.lulzimg.com/5224090594.png

You could try my kernel with SDK 2.6! On my 6950 I get a 30 MH/s increase in Phoenix 1.7 over current CGMINER 2.0.8. I guess some of Phateus optimisations don't work anymore, because AMD updated the OpenCL compiler to generate better performing code.

Dia

*COUGH* perhaps if someone posted "my kernal" We could learn what the fuck "my kernal" is

WELL!, Nice work, However... Simply switching to your kernal, Gives me back the Mhash i "lost" But at the same time, Gives me back the CPU load bug....
Not pleasant... Also noticed using GUIminer w/Phoenix -k phatk AGGRESSION=12 FASTLOOP=false VECTORS2 WORKSIZE=128

Since the CPU pegs at 100%, When i attempt to run a 2nd GPU on the same 100% pegged core, I lose 20mhash on both cards. This does not happen with the regular kernal.
So if i want to run two of my gpu's using this Nice -k phatk AGGRESSION=12 FASTLOOP=false VECTORS2 WORKSIZE=128 then i need to PEG two cores at 100%, HOTHOTHOT NO THANKS

Just thought "when was the last time i installed an SDK, Dont they come with the drivers?" *clickclickclick* Oh that might be my problem.... *clickclick* installind SDK APP 2.6

Just installed 2.6sdk, No differance.
Guiminer+Phoenix+Phatk=CPU peg 100%, But the Mhash "loss" is gone, So this is like rolling back
Guiminer+Poclbm=10mhash loss percard, But no CPU 100% bug

well lol i see no cpu bug anymore with cgminer and i see hardly any loss of performance but i start cgminer simple on my 5870, 5970 computers like :

cgminer v 2 w 128 -I D --temp-cutoff 95 --auto-gpu --queue 1 --gpu-powertune 20 --gpu-engine 850-920 --gpu-memclock 300 -o http://192.168.178.5:9332 -u user -p pass

And it runs like charm when i set it to i7 it really keeps a steady average, but i need the machine for other things then mining as well ;)
Setting intensity at higher levels has a negative effect and i do not see any difference using them.
With I D it runs at 420 Mh with i6,i7 or i8 its all the same just when i use other programs i get some loss of performance as expected with -I D

Edit: After some testing they go off again ... untill BTC reached $ 30 again

Sorry for not posting a direct link - https://bitcointalk.org/index.php?topic=25860.0
Well I think the 100% CPU usage is not a fault in AMDs drivers but it comes from how I process the OpenCL buffer, which holds nonces.
To speed up kernel execution I removed control-flow (if-statements) from the kernel, but now check the whole buffer for valid nonces, even if there are none ... this leads (my guess) to an endless processing loop in Phoenix, which is the drawback of the kernel changes.

Dia

... so just don't use it.

Dia

+1

Just as a data point, cgminer does not do this. It does not check the whole buffer.

I get 280 instead of 334mhs with 12.1 (5850 at 825mhz 300ram poclbm -v -w256)

That is not some small drop ... this is 20%,  I was expecting ~5% and that it was worth it for the CPU bug.

try with different settings cause when i use -v w256 i have a lot less performance
setting it to -v2 -w 128 should fit better on a 5850

Hi Con,

That is correct, but CGMINER uses Phatk2, which seems currently to not work that well with SDK / runtime 2.6.
So it's again a trial and error to get the best tradeoff.

Dia

Edit: I found out that a Worksize of 64 with Phoenix works great now, this was much slower before 2.6.

For those who got less hash with 12.1 with cgminer

Try worksize 64 with vectors 4

That's interesting because each GPU will report what its "preferred vector size" is, and often it comes out to 4, yet despite that, virtually all GPUs had much better performance (with the older SDK) and 2 vectors. Perhaps they live up to their promise now?

Nope, preferred vector size does not always mean "best performance". Wider vectors mean more GPRs used and the more GPRs you use, the less wavefronts you can schedule on a CU, thus occupancy goes down. Also, on VLIW5 hardware, uint4 vectors are not optimal, it can happen that there are no 5 non-dependent instructions to fill the whole VLIW bundle. That depends on your code.

For example, with hash cracking you might end up with uint8 being much better than uint4 for kernels like the MD5 or NTLM one as ALUPacking goes up and the number of used GPRs is 10-20 at most. On the other hand, more complex algorithms like SHA512 are much better with uint2 or even a scalar implementation as the number of used GPRs greatly hampers the occupancy. With memory-intensive kernels like DES ones (thanks god bitcoin is not one), occupancy becomes even more important as more concurrency means memory access  latencies are more easily "hidden".

Back to uint2, problem with it is that it's even worse at utilizing all the slots in the VLIW bundle and your ALUPacking just always sucks. However, generally speaking with most bitcoin kernels it's a tradeoff worth having as bad occupancy in that particular case is worse than bad ALUPacking.

uint3 should provide a better balance between those, but it was broken in pre-2.6 APP SDK releases. Now they fixed it and I am curious about results...


PS as of why suddenly uint4 started performing better, it could be either that they iimproved scheduling or that they have improved the backend compiler to pack instructions better with uint4 / worse with uint2. It could be actually both.

Oh btw there are two different versions of 12.1 preview

The one I tried is this:
http://developer.amd.com/Downloads/OpenCL1.2-Static-Cplus-preview-drivers-Windows.exe

It has a newer openCL than the other 12.1 preview

I implemented uint3 quite a few months ago, but it was bugged (like you said) ... since 2.6 I'm able to compile my kernel via KernelAnalyzer and get no errors (the uint3 kernel is much longer by the way). But I guess I did something wrong in the init.py from Phoenix, which I can't solve by myself ... Phoenix crashes if kernel is started. Are you skilled enough to take a look at it?

Dia

Not quite, python is not among my strong sides. I may rewrite my miner though, just for the experiment. Anyway I have more important projects right now.

Confirming a 20% drop in hashing performance across a variety of cards using 11.12.

Fuck.

Time to roll back to older drivers.

I hate ATI so damned much sometimes.

Edit: Using phoenix/phatk

Could you try this one (https://bitcointalk.org/index.php?topic=25860) with Phoenix 1.7.1 and report your results?

Dia

10% drop across all my cards. But no more cpu bug, Cant tell wich is worse.

So can anyone bring there hash rate back to normal with the cat 12.1? I love it's gaming performance, But im losing 10% of my speed so that i can save 17% of my cpu usage... I'd rather not burn out one of my cores, Atop of that it's a waste of power.

I confirm a 30+ Mhs drop on both a 5870 and a 5830 with both CataCLYSM 11.12 and 12.1 compared with 11.6.
What is MUCH WORSE is that you cannot revert back effectively: uninstalling and driver sweeping before rebooting and reinstalling 11.6 will NOT return your previous performance.
Luckily, I had a fresh system backup made just yesterday that I could restore no prob.
Now I am back to mining happily with my 5870@970/300 hashing at 440Mhs+ and a 5830@940/300 hashing at 300Mhs+ with Phoenix 1.7 on Xp.

We really REALLY need to optomise these new drivers...
Some of us still play videogames and mine coins y'know... It's not just all Deticated miners running

mining works good for me no cpu usage and still good perf but i been crashing in games on so im going back to try 11.12.

...Wtf...
12.1 i've never crashed yet...
Im LOVING these drivers. but the lack of 10% mining speed really hurts

It would crash in dead island 1-4hrs at random 11.12 does not and it still has 0 % cpu usage but the lower mining speed to.

Ohhh yeahhh.... That dead island thing...

What that is is the total number of Dead Bodies not being Properly counted/Dumped In/out of your ram, It's basically a memory leak that floods your video memory until you crash.

Like when you change zones, The dead bodies are still in your ram.. But they're not there anymore.. *sigh*
i S'pose the fact that i Didnt like that game, Have large Vram and never played it for more than 4hrs in one sitting, Would equal that i dont crash on it

could be i run a 5970 and 5870 in crossfire game runs great other then the bsod i was getting on the 12.1 11.12 are nice to other then i cant add manual crossfire profiles with it like 12.1p.  I just installed sdk 2.1 with my 11.12 drivers to get bettter speeds mining and still 0% cpu usage and can still game.

im running 11.12 and sdk 2.1 right now no cpu bug i uninstalled the new sdk and installed 2.1 no problems now.  The sdk should not affect games mine play fine i dont play bf3 best thing you can do it try it and see.
http://i.lulzimg.com/cbfcdb738d.png

I'm using 12.1 preview drivers, on Windows 7 Home Premium 64 bit.  With Phoenix 1.72 using pocblm Aggression 7 Worksize 64 I get ~433 mhash on my Stock HD 6970.  That's about 40 mhash better than it ever got before and there is at most 1-2% CPU useage when the rest of the system is at Idle.

I wonder how much power it saves not having 100% cpu usage on one core. Probably worth more in savings than the 8-10mh/s drop per card.

How the heck are you running guiminer with sdk 2.1? Whenever I run guiminer with sdk 2.1 I get an error saying no opencl devices are installed. I keep seeing really good hashrates with 2.1 but I really want to use guiminer too and you somehow managed it.

show me a screenshot of your gpu caps viewer of the opencl page while guiminer is mining. you say you uninstalled your previous sdk, but thats kind of suspicious seeing that almost all posts concerning 2.1 sdk + guiminer says it wont work.

EDIT: NEVER MIND! I found a way to use guiminer + 2.1 SDK. Guiminer doesn't like 2.1 SDK by itself. If you have ANOTHER sdk installed ALONG SIDE of 2.1 though, then you will have 2 opencl platforms to choose from. One of the platforms will be 2.1 and guiminer and phoenix will happily mine away faster!

Well. That because you have a 69xx, and Phatkernal variations running with sdk 2.6 depend on the mem clock being "close to stock" Such as 1000, or 1150
People without a 69xx that want to have a Lowered memory clock cant use Phatkernal variations >.< (with sdk 2.6)

Thanks for this...

This is awesome. Thanks for this. Hashing rate increased from ~120 to ~140 (cgminer 2.1.2, dynamic intensity, Radeon 5750, stock clocks, memory 150MHz). Small donation sent  :D

Wow THANKYOU! this saved my butt because i had copied the wrong files manually, even a reinstall did not fix it.

Flipped you a quarter ;^)

This fix worked great for me on Xp and Xp64, but not on Seven32. I just get some file version error. Anyone has any idea?

Windows 7 and Windows XP require different drivers ...

Suupperrr weirrdd...
Works on xp32, xp64 win7x64 but totally fails on win7x86

Anyone from AMD/ATI ever pipe in on the bitcoin mining arguments/testing/improvements in hardware/software ?

Way to necro the thread dude... You totally could've posted that in a different thread

Could have, just didn't want to