52s at stock speed is not bad. What do you get running a single WU when overclocked? My 5870 does about 59s per single WU overclocked. 74s is my 5970 but that is not OC much due to being on stock cooler.

52s at stock speed is not bad. What do you get running a single WU when overclocked? My 5870 does about 59s per single WU overclocked. 74s is my 5970 but that is not OC much due to being on stock cooler.

Bare in mind the timings I give are running an OpenCL application which puts a speed penalty on the 79XX. I have two 7970s in a mid size LianLi case, it has very good airflow as I fitted pressure fans, so bare that in mind re temperatures, as the pressure fans help a lot. The cards are reference model ASUS, stock cooler.

- On cold startup a single card one WU, no app_info is 52 seconds

- After around 30mins that 52 seconds goes down to around 49-52 seconds as the effect of cache, predictive software etc licks in.

- On cold start of two WUs per card running app_info, it will start at 45 secs per WU, then go down to 43-44 secs per WU after 30 mins or so on 1245 GPU o/c.

Once settled down, the cards when running two per card (my normal mode at present) are at 76 and 78 degrees. They are running at 1.23v (default is 1.174v, card max is 1.3v), 1245 GPU / 1375 Memory. I've seen o/c sessions verified at 1260 GPU on air and doing a benchmark, ie stable. I am finding very faint blemish on minute parts of the screen at 1250 GPU, so backed off to 1245, I'll probably settle at 1240 for safety long term .

As to a single WU o/c - I havent really done one, but based on the two per card results I would expect around 43-44 seconds a WU with 1240 GPU.

Conventional wisdom has it that there is a large hit if running Crossfire/SLI. That was true for sure in the 5XXX genre, but changed in 6XXX, and now with 7XXX scaling is real good, its around 97/98 %, so in reality terms at MW you loose about a second on each WU by using crossfire.

Power is excellent, as would be expected as its a 28nm card. Its around 210w going full stretch inside the driver tweaking, and can go up to around around 250-260w when its going full tilt and overvolted. Guru3D have done three excellent reviews on 7970 from different angles, and they published today three reviews on the 7950 which is now out there.

Whilst its very good, keep perspective, NVidia yet to deploy their first 28nm offering and they should be at these levels, no reason why not. Timning may be an issue with some as they will be probably mid Summer before all theirs are out. However, dont count NVidia out, they were a good harware company circa 2005/2006 but went haywire. 5XX has helped, but now is their defining moment, they got to get 6XX right .... time will tell, but dont write them off, I have a feeling that for the first time in 5 years they may well bring out a real good one, hope so, it push all prices down :)

If I missed anything you need please yell, not a problem.

Regards
Zy

why does my 7970 not run any tasks, i have a 6990 and a 6970 anmd just added a 7970 and the 7970 is the only card recognized and it will not run any of the tasks assigned to it

why does my 7970 not run any tasks, i have a 6990 and a 6970 anmd just added a 7970 and the 7970 is the only card recognized and it will not run any of the tasks assigned to it

The 7970 expects to run an Open CL app. Look here:

http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=2741&nowrap=true#52725

.... and a few posts either side of that, its a chat with Steve getting his going. The key thing is you must manually install a few things for now, and must use an app_info file as the application is an NVidia compiler OpenCL app, and for the moment you have to explicitly tell the server you need it, else it doesnt know.

To get the driver you need (a special one for 7XXX - dont use regular AMD drivers, go here to download:

http://downloads.guru3d.com/download.php?det=2844

For the install instructions and other bits you need, look at this link, and go to Win7 64 bit, and select the bottom one of the three (OpenCL CUDA app 0.82), its a download link for a zip file. [and say thank you to Arkayn for putting it together on his site with his usual quiet efficient support he gives us all]

http://www.arkayn.us/forum/index.php?action=downloads;cat=11

In there is all you need. If have issues, please post, dont reinvent the wheel - there's lots at the end of the @7970 Dont Run@ thread which will help, but please, ask, dont waste time reinventing wheels. Sounds a mouthful, but there is not much to do in truth - just wordy wordy explaining it.

Regards
Zy

There is a new Version of the ASUS GPU Tweak Utility out - ASUS GPU Tweak V2.0.3, it is very good, and well worth a download. The link below is for the utility on 7970 cards. It may work on other cards despite the title on the main screen, I have no idea.

They have also folded into it GPU-Z. The result is a neat collection of tools, combining some of the Afterburner and GPU-Z functionality. The tool set is not as extensive as Afterburner, but its close .... real nice. Not had a chance to test thoroughly yet, but so far so good.

It's supposed to work with all cards. Tried it on one machine. Even in advanced mode it detected only 1 GPU out of 2. Any change in settings caused a reboot. Exiting the program caused a reboot. When it ran the Dimes client wouldn't. Uninstalling worked at least. Not even close to ready for prime time. Stick to Afterburner. BTW Zy, could you also PM me with your AB hints you mentioned above? Thanks!

In the end I decided to refuse my 7970 delivery and send it back. It's not worth spending over £400

Yeah, buying it for BOINC is not worth it right now. You can pull the trigger whenever the software is ready, and at that point it will certainly not be at £400 any more.

MrS

---

Scanning for our furry friends since Jan 2002

Tried it on one machine. Even in advanced mode it detected only 1 GPU out of 2. Any change in settings caused a reboot. Exiting the program caused a reboot. When it ran the Dimes client wouldn't. Uninstalling worked at least. Not even close to ready for prime time. Stick to Afterburner.

Wow .... after seeing that lot ... I'm with the guy holding the software axe rofl :)

Regards
Zy

Hi,

2x7970@1080-1700 with cat 11.3 prewhql :
one wu in 38sec for each so two wu in 38sec
@stock each wu last 44sec

You mean Cat 12.3 preview? That would be way faster than with previous drivers, isn't it?

MrS

---

Scanning for our furry friends since Jan 2002

this driver http://support.amd.com/us/kbarticles/Pages/catalyst122precertifieddriver.aspx
and it says 12.2 pre certified driver but in ccc it says 12.3
I didn't try with previous driver
and one question : is 38 sec fast ? i don't have a reference to compare

edit : when i see this http://forums.anandtech.com/archive/index.php/t-2064356.html, it seems fast but is the current wu comparable to those of march 2011 ?

excuse me for my english, it isn't my native tongue

...and one question : is 38 sec fast ? i don't have a reference to compare

For 7970 at those settings its fine. One thing to remember about MW - you dont need high memory. Suggest you turn up the GPU in CCC to 1125, and memory down to 1250. Any setting below 1375 for memory on the 7970 is only a "target" setting, the voltage controller will only go to that if it thinks it can do so safely. Memory settings are not real time with the type of voltage regulator on the card.

Depending on the 7970 card, its possible you may have to bring the GPU down a little, it may not run in CCC at 1125, should fine, but dont be surprise if you do need to bring it down a little. You must ensure the power slider is at maximum (+20%) when o/c GPU in CCC. Without that +20% you will have problems.

The card and the MW WU will go faster with GPU at 1200, however you need to over-volt to achieve that, do not attempt overvolting without someone to help you if have not done it before - you will burn the card otherwise. After a while at 1200, it can go to 58-60 secs for two on a card ie +/- 30 secs per WU. For the moment, +/- 30 secs is as far as it will go. The card has more legs in it, but thats for the future and another thread, right now, its as fast as it will go (if you want the card to survive long term .... ).

Regards
Zy

Without that +20% you will have problems.

Has anyone actually tested this? I highly doubt you'll draw more than 250 W running MW, even overclocked, as long as you stay at or below 1.175 V (stock). In games the card hardly draws more than 200 W.

Plus I'm running my HD6950 overclocked and with unlocked shaders, yet don't need to increase the default Power Tune limit of 200 W. Heck, the entire machine is drawing < 300 W from the wall with an i7 @ 4 GHz and full BOINC load.

Edit: yes, 38s is fast! Makes me a little jealous ;)

MrS

---

Scanning for our furry friends since Jan 2002

I highly doubt you'll draw more than 250 W running MW, even overclocked, as long as you stay at or below 1.175 V (stock). In games the card hardly draws more than 200 W.

Its whats "under the hood", away from prying eyes (normally) thats the problem. All graphics cards vary in the voltages used not only by design and the type of voltage regulator used, but used by outcomes in the Fabrication Plant as a consequence of the fabrication process. The real base frequency is not what you see on a screen as reported (eg 1.174v is not what it seems).

The 7970 in reality can operate at four different base voltages, these base voltages are adjusted internally to 1.174v on start up of the card, hemce the reason it seems that 1.174v is the base voltage - it isnt. For the 7970, there are in fact four different standards of base voltages, depending how the card is Binned at the Fabrication plant:

1.175V, 1.025V 1.1125V or 1.05V

Now when you reflect on the useage of the Power Slider - which is not to supply additional power, but its there to raise the card TDP to allow additional power - together with the fact that you are playing with different base voltage levels, it comes clear why some people have reported issues going to 1125 in CCC with the 7970, even though the power slider was at +20%. The power slider is not the news bit, it changes TDP nothing more, the key factor is knowing the voltage levels applying to the 7970 compared to other card types beit AMD or NVidia (who also have varying voltage levels). In the link below is how to discover the card (actual) base voltage (not the one shown on screen)

Rather then regurgitate the lot, the thread below - 6 weeks old - is a key read on this. Start from post one, and work through it. Halfway through you will see Unwinder steer things and a series of tests being carried out. He wanted to verify a couple of things re the 7970.

(For those not aware, Unwinder is the Designer of Riva Tuner, and the Designer and Architect of MSI Afterburner, which has Riva Tuner as its Core engine - what he doesnt know about voltage regulators, can be fitted on the back of a postage stamp)

http://forum.guru3d.com/showthread.php?t=356803

Regards
Zy

I looked through this thread for "Halfway through you will see Unwinder steer things and a series of tests being carried out. He wanted to verify a couple of things re the 7970."

But didn't find it. Am I missing something?

The tests and final pointer to the voltages start from post number 117, the posts prior to that need reading to get the full impact of whats happening.

http://forum.guru3d.com/showpost.php?p=4221583&postcount=117

Regards
Zy

To summarize that thread:

- Tahiti is available with 4 load voltage bins (as you listed)
- current Afterburner shows the voltage correctly, whereas other tools might just get 1.175 V from a database
- the chips are binned by leakage and, as it seems, high leakage chips get lower voltages to keep their power consumption approximately in line with the others
- under load the real GPU voltage drops, just as for CPUs.. but this doesn't matter for us
- most cards are 1.175 V default

I take from this that the "typical power consumption" I listed previously is still correct, as the lower voltage parts make up for that with their higher leakage. They would consume a bit more power if run at the same voltage, though.

And I still stand by "if the card typically draws ~200 W, there shouldn't be a need to allow 300 W instad of 250 W".

MrS

---

Scanning for our furry friends since Jan 2002

For sure, power draw never been in doubt .... but if a user has a lower voltage card, then issues will occur trying to go 1125 (originally hoping CCC will do it) without a voltage increase. If they are not aware of that, lots of fruitless searching for an answer that does not exist will occur, as their perception will be one of puzzlement that CCC is not working - it is, just that they are the lucky ones with a more overclockable card, and if on the lower voltage, its going nowhere to 1125 or beyond without voltage increases

Unless they understand the actual card voltage setup, which the thread gives them.

Regards
Zy

Since the most recent update I can no longer get my 7979 to use 100% of the GPU

If I'm using app_info.xml does that override the web settings? If so what do I need to put in the app_info.xml to get the gpu crunching full power again?

Running drivers after the release drivers but before the 12.2 drivers :)

---

Which up date - an AMD driver ? If so, I'd try still using the release day drivers as the changes to the driver sets so far have been gamer related not Compute - obvious dilema if your a gamer as well .... but there ya go :)

What utilisation figure do you have in CCC ? If its 95%+ thats pretty well it, 97-99% is the norm on 7970s. Significantly less than that circa 85-90% then there maybe an issue lurking. As to what hard to say with no more symptoms.

An app_info will allow 100% useage running two WUs per GPU. It can alsohide issues such as you described, so another dilema sometimes. Below is an app_info that will put two WU per gpu on a card. Note the cmd line frequency value. Thats a low 10 (old default was 30, new default is 60). The 7970s respond quickly switching to the display screen to service it, so a low value usually works ok. If the screen lags horribly, step up the value in steps of 5 until you are happy with screen response - should be ok, but step up slowly like that if issues.

Dont forget to clear cache before using the app_info.xml, and to restart BAM after inserting it into the Milkyway program directory. Use Notepad to create the app_info.xml to prevent hidden charactors being inserted making it un-useable.

<app_info>
<app>
<name>milkyway</name>
</app>
<file_info>
<name>milkyway_separation_1.02_windows_x86_64__opencl_amd_ati.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>milkyway</app_name>
<version_num>102</version_num>
<flops>1.0e11</flops>
<avg_ncpus>0.05</avg_ncpus>
<max_ncpus>1</max_ncpus>
<plan_class>ati14ati</plan_class>
<coproc>
<type>ATI</type>
<count>0.5</count>
</coproc>
<cmdline>--gpu-target-frequency 10 --gpu-disable-checkpointing</cmdline>
<file_ref>
<file_name>milkyway_separation_1.02_windows_x86_64__opencl_amd_ati.exe</file_name>
<main_program/>
</file_ref>
</app_version>
</app_info>

Regards
Zy

Which up date - an AMD driver ? If so, I'd try still using the release day drivers as the changes to the driver sets so far have been gamer related not Compute - obvious dilema if your a gamer as well .... but there ya go :)

8.921.2 RC11

What utilisation figure do you have in CCC ? If its 95%+ thats pretty well it, 97-99% is the norm on 7970s. Significantly less than that circa 85-90% then there maybe an issue lurking. As to what hard to say with no more symptoms.

75-85%

Before the default frequency change I was getting 99% on the card regardless of running one or two instances. I also found no gain by running two instances as I was getting 36 sec/wu just running one at a time, and similar speed with two at a time but much higher temps/noise.

An app_info will allow 100% useage running two WUs per GPU. It can alsohide issues such as you described, so another dilema sometimes. Below is an app_info that will put two WU per gpu on a card. Note the cmd line frequency value. Thats a low 10 (old default was 30, new default is 60). The 7970s respond quickly switching to the display screen to service it, so a low value usually works ok. If the screen lags horribly, step up the value in steps of 5 until you are happy with screen response - should be ok, but step up slowly like that if issues.

I changed the web frequency to 20 but it didn't seem to make any difference, does having an app_info.xml stop the client using the web pref settings?

Dont forget to clear cache before using the app_info.xml, and to restart BAM after inserting it into the Milkyway program directory. Use Notepad to create the app_info.xml to prevent hidden charactors being inserted making it un-useable.

<cmdline>--gpu-target-frequency 10 --gpu-disable-checkpointing</cmdline>

Regards
Zy

Great I'll try both an app_info with the --gpu-target-frequency set and I'll try without the app_info (as I'm only running one task at a time)

Haven't noticed any desktop issues but then its probably because its ignoring the web settings as the app_info is there. :D

---