What catalyst and sdk drivers are you using for 7970 (windows)

[gateway]

Just taking a poll as to see what people are running..

Currently I have C 12.8, which I believe comes with sdk 2.8

I'm running cgminer on a sapphire 7970 oc'ed

settings:
gpu = 1183
mem = 1040
v = 1243
boost 20

temp is around 72, with fan speed of about 83

pulling in 702-698~

[1713241691]

1713241691

[1713241691]

1713241691

[1713241691]

1713241691

[crazyates]

13.1 which comes with 2.8

Stock 1100MHz, with 950 memory, 1.256V, Powertune set to default, and 80% fan, 66C @ 661MH/s.

[gateway]

13.1 which comes with 2.8

Stock 1100MHz, with 950 memory, 1.256V, Powertune set to default, and 80% fan, 66C @ 661MH/s.

nice on the temp.. what card are you using and do you recommend it..

At that voltage you should be able to push to 1040 maybe.. not sure what settings you tried.

[crazyates]

13.1 which comes with 2.8

Stock 1100MHz, with 950 memory, 1.256V, Powertune set to default, and 80% fan, 66C @ 661MH/s.

nice on the temp.. what card are you using and do you recommend it..

At that voltage you should be able to push to 1040 maybe.. not sure what settings you tried.

I'm using the Gigabyte TriForce one. Giant heatsink, with 3 fans. The VRM temps are a lil higher than a stock, but that's to be expected, and not bad. I can OC to ~1200MHz for 700+MH/s, but it uses a lot more power. I used to undervolt at 1050mV for 1050MHz, and the temps were like 50C @ 50% fan. It was nice.

[iCEBREAKER]

Default Powertune?

My 7970s really like +20.

[crazyates]

Default Powertune?

My 7970s really like +20.

Ya I've always left the powertune alone. I think the Gigabyte I have runs with a little higher voltage to accommodate the stock 1100Mhz, so I've never had any issues. Should I try upping it to 20?

[iCEBREAKER]

Default Powertune?

My 7970s really like +20.

Ya I've always left the powertune alone. I think the Gigabyte I have runs with a little higher voltage to accommodate the stock 1100Mhz, so I've never had any issues. Should I try upping it to 20?

Yes, depending on your local conditions (temp and $/kwh).  It will run hotter and draw more power but give you a nice bump in MH/s.

[crazyates]

Default Powertune?

My 7970s really like +20.

Ya I've always left the powertune alone. I think the Gigabyte I have runs with a little higher voltage to accommodate the stock 1100Mhz, so I've never had any issues. Should I try upping it to 20?

Yes, depending on your local conditions (temp and $/kwh).  It will run hotter and draw more power but give you a nice bump in MH/s.

Ya I don't notice any difference using CGMiner to change the PowerTune anywhere from -20 to +20. I have it at 1100/950/1.256V/80% fan. Stays the same 661MH/s.

[scifimike12]

Catalyst 13.3 Beta 3
OpenCL: 10.0.1124.2

7970
1000/300 (1.050v) ~ 601 MH/s
Max temp: 37°C

1010/300 (0.987v) ~ 608 MH/s
Max temp: 37°C

1050/300 (1.010v) ~ 635 MH/s
Max temp: 38°C

7950
1100/300 (1.024v) ~ 577 MH/s
Max temp: 38°C

[DrG]

Win 7 x64 SP1, no updates.  Catalyst 12.8 with all the crap installed and default SDK

Pair of Diamond 7970 voltage unlocked but running stock voltage
1175 Core/200 mem/1.112 v (stock) - 703MH/s per card - cards range 68C to 86C.

The rooms runs between 80F-105F depending on if AC is running - we can't all live inside a fridge like Tacotime  

[PatMan]

1070/1200
stock volt
max temp 72
700MH/s steady

1200 memclock seemed to be the sweet spot, uses slightly more power but seems worth it for the extra hash rate. Cards are happy.

[crazyates]

1070/1200
stock volt
max temp 72
700MH/s steady

1200 memclock seemed to be the sweet spot, uses slightly more power but seems worth it for the extra hash rate. Cards are happy.

You either need to post your OS, driver, SDK, miner version, and all of your command line arguments/settings, or I'm gonna call BS. Everyone else needs close to a 1200 core speed to get to 700MH/s.

[PatMan]

1070/1200
stock volt
max temp 72
700MH/s steady

1200 memclock seemed to be the sweet spot, uses slightly more power but seems worth it for the extra hash rate. Cards are happy.

You either need to post your OS, driver, SDK, miner version, and all of your command line arguments/settings, or I'm gonna call BS. Everyone else needs close to a 1200 core speed to get to 700MH/s.

Ah! You found a typo - well spotted.....

1170/1200, not 1070

As for the rest:

Xubuntu 12.04
Driver 12.3
SDK 2.6
cgminer 2.11.3

[gateway]

Has anyone done tests between windows and ubuntu.. as to whats more stable or better hash rates.. obviously windows is easier to set up.

I'm also seeing people using the latest cat drivers 13.1 reporting good results.

So also whats the deal with memory..

some people say to keep mem 150 lower than gpu clock.. see memdiff setting in cgminer..

My by default has a gpu clock at 1000 and memory at 1450.. just sorta kinda trying to figure out a good combo.. of course this requires some serious babysitting to see when it goes down.

[crazyates]

So also whats the deal with memory..

some people say to keep mem 150 lower than gpu clock.. see memdiff setting in cgminer..

My by default has a gpu clock at 1000 and memory at 1450.. just sorta kinda trying to figure out a good combo.. of course this requires some serious babysitting to see when it goes down.

Lower memory is better for temps and power usage. Higher memory does not affect hashrate. CGMiner uses the standard ADL, which cannot clock the memory more than 150 below the core engine speed. Other tools like MSI AB can. I ran my 7970 at 1050/310 for quite a while, when undervolting was more profitable.

[ocminer]

So also whats the deal with memory..

some people say to keep mem 150 lower than gpu clock.. see memdiff setting in cgminer..

My by default has a gpu clock at 1000 and memory at 1450.. just sorta kinda trying to figure out a good combo.. of course this requires some serious babysitting to see when it goes down.

Lower memory is better for temps and power usage. Higher memory does not affect hashrate. CGMiner uses the standard ADL, which cannot clock the memory more than 150 below the core engine speed. Other tools like MSI AB can. I ran my 7970 at 1050/310 for quite a while, when undervolting was more profitable.

Where can I get the "non-standard" ADL ? (for linux)

[crazyates]

So also whats the deal with memory..

some people say to keep mem 150 lower than gpu clock.. see memdiff setting in cgminer..

My by default has a gpu clock at 1000 and memory at 1450.. just sorta kinda trying to figure out a good combo.. of course this requires some serious babysitting to see when it goes down.

Lower memory is better for temps and power usage. Higher memory does not affect hashrate. CGMiner uses the standard ADL, which cannot clock the memory more than 150 below the core engine speed. Other tools like MSI AB can. I ran my 7970 at 1050/310 for quite a while, when undervolting was more profitable.

Where can I get the "non-standard" ADL ? (for linux)

Sorry, I may not have been clear. There is no "non-standard" ADL. The only reason why windows users can clock their memory lower is through program like MSI AB, which use custom and unofficial overclocking methods. IDK if these methods can even be duplicated in Linux, but even if they are, there are no programs out there that do it. You will just have to stick with memory at the medium speeds.

For example: My 7970 stock is 1100/1500. I used to run it at 1100:310, but I hated MSI AB, so now I'm running clocking it thru CGMiner, and it's at 1100/950. It's running fine, and there are no real side effects besides slightly higher power and temps, but nothing crazy. I still consider it worth it.

[ocminer]

So also whats the deal with memory..

some people say to keep mem 150 lower than gpu clock.. see memdiff setting in cgminer..

My by default has a gpu clock at 1000 and memory at 1450.. just sorta kinda trying to figure out a good combo.. of course this requires some serious babysitting to see when it goes down.

Lower memory is better for temps and power usage. Higher memory does not affect hashrate. CGMiner uses the standard ADL, which cannot clock the memory more than 150 below the core engine speed. Other tools like MSI AB can. I ran my 7970 at 1050/310 for quite a while, when undervolting was more profitable.

Where can I get the "non-standard" ADL ? (for linux)

Sorry, I may not have been clear. There is no "non-standard" ADL. The only reason why windows users can clock their memory lower is through program like MSI AB, which use custom and unofficial overclocking methods. IDK if these methods can even be duplicated in Linux, but even if they are, there are no programs out there that do it. You will just have to stick with memory at the medium speeds.

For example: My 7970 stock is 1100/1500. I used to run it at 1100:310, but I hated MSI AB, so now I'm running clocking it thru CGMiner, and it's at 1100/950. It's running fine, and there are no real side effects besides slightly higher power and temps, but nothing crazy. I still consider it worth it.

When I try to lower my memory on my 7970 XFX Ghost cgminer freezes my whole System and I have to reboot. Even a "even underclock (1000/1000) is not possible.. It fails immediately when I try it on startup with cgminer but works *sometimes* when I do it out of a running instance ... very odd.

Linux ubuntu 12.10 with latest beta driver and SDK 2.8 ..

Never had so much trouble than with these 7970 cards and I'm mining quite long :/ They're really not worth the hassle.. I'll probably stick to some 5870/5970 .. but they are starting to become really rare..

[PatMan]

So also whats the deal with memory..

some people say to keep mem 150 lower than gpu clock.. see memdiff setting in cgminer..

My by default has a gpu clock at 1000 and memory at 1450.. just sorta kinda trying to figure out a good combo.. of course this requires some serious babysitting to see when it goes down.

Lower memory is better for temps and power usage. Higher memory does not affect hashrate. CGMiner uses the standard ADL, which cannot clock the memory more than 150 below the core engine speed. Other tools like MSI AB can. I ran my 7970 at 1050/310 for quite a while, when undervolting was more profitable.

Where can I get the "non-standard" ADL ? (for linux)

Sorry, I may not have been clear. There is no "non-standard" ADL. The only reason why windows users can clock their memory lower is through program like MSI AB, which use custom and unofficial overclocking methods. IDK if these methods can even be duplicated in Linux, but even if they are, there are no programs out there that do it. You will just have to stick with memory at the medium speeds.

For example: My 7970 stock is 1100/1500. I used to run it at 1100:310, but I hated MSI AB, so now I'm running clocking it thru CGMiner, and it's at 1100/950. It's running fine, and there are no real side effects besides slightly higher power and temps, but nothing crazy. I still consider it worth it.

When I try to lower my memory on my 7970 XFX Ghost cgminer freezes my whole System and I have to reboot. Even a "even underclock (1000/1000) is not possible.. It fails immediately when I try it on startup with cgminer but works *sometimes* when I do it out of a running instance ... very odd.

Linux ubuntu 12.10 with latest beta driver and SDK 2.8 ..

Never had so much trouble than with these 7970 cards and I'm mining quite long :/ They're really not worth the hassle.. I'll probably stick to some 5870/5970 .. but they are starting to become really rare..

Strange, they run brilliantly on 12.04. Sounds like a driver/sdk problem though tbh, not sure what the best combo is for 12.10 as I've not used it much, sorry. I presume you've tried a few different combos?

[ssateneth]

Have an update. Figured I'd bump.

Reference OEM 7970, 13.3 beta, SDK that came with it. CGMiner 2.11.4, Diablo kernel, Worksize 256, Default vectors (1), intensity 11, 2 GPU Threads (default). Don't need to specify any other options aside from pool and GPU speeds. 1185 Core, 300 memory.

713.2 MHash/sec, or ~0.60186 MHash/MHz.

Running with 1 GPU thread was about 700 on the nose. Also, I have not experimented with all the combinations quite yet.

Edit: On another 7970, I was able to get slightly different core clocks, namely 1220/300, and I tested other kernel combinations and let them hash out for about 4-8 hours to get a good average that wouldn't fluctuate. Increasing vectors decreased hashrate significantly. Increasing threads beyond 2 either decreased hashrate significantly or caused a driver crash and recover. The default GPU threads (2) and default vectors (1) seem to be optimal.

730.1 MHash/sec (0.59844 MHash/MHz) with poclbm kernel and worksize=256
731.3 MHash/sec (0.59943 MHash/MHz) with poclbm kernel and worksize=128
728.1 MHash/sec (0.59680 MHash/MHz) with poclbm kernel and worksize=64 (This is the default setting.)
730.8 MHash/sec (0.59902 MHash/MHz) with diablo kernel and worksize=256
731.7 MHash/sec (0.59975 MHash/MHz) with diablo kernel and worksize=128
726.6 MHash/sec (0.59557 MHash/MHz) with diablo kernel and worksize=64

The hashrate per MHz seems to be slightly lower, probably because of a lower RAM:Core ratio. I will increase it slightly to see if I can eek out a slightly better MHash/MHz.