Message boards :
Number crunching :
79XX Dont Run
Message board moderation
Previous · 1 · 2 · 3 · 4 · 5 · 6 . . . 7 · Next
Author | Message |
---|---|
![]() ![]() Send message Joined: 15 Jul 08 Posts: 383 Credit: 729,293,740 RAC: 0 ![]() ![]() |
None - sorry - its a generalised utility that covers essentials, a throw away vender bribe rofl - so its kind of as is. Cant think of any workaround it might have for that. Unfortunately makes it pretty useless for multiple GPU setups, even setting a fixed fan speed on one card resets the other to automatic and the temp skyrockets. Maybe either I'm missing something obvious or it'll get fixed someday. Afterburner works pretty good for the most part and controls NV cards as well. |
![]() Send message Joined: 2 Sep 07 Posts: 24 Credit: 26,244,596 RAC: 0 ![]() ![]() |
I haven't had issues in the pat with TRIXX setting custom fan profiles on multiple cards... but I don't have a dual GPU setup atm |
![]() Send message Joined: 1 Sep 08 Posts: 204 Credit: 219,354,537 RAC: 0 ![]() ![]() |
But in Afterburner you can't downclock the RAM by much.. as it would be efficient for MW. I even asked them some time ago to somehow let people adjust the lower clock limit, to make their tool more useful. They said, it wouldn't be in the spirit of Afterburner.. MrS Scanning for our furry friends since Jan 2002 |
![]() ![]() Send message Joined: 15 Jul 08 Posts: 383 Credit: 729,293,740 RAC: 0 ![]() ![]() |
But in Afterburner you can't downclock the RAM by much.. as it would be efficient for MW. I even asked them some time ago to somehow let people adjust the lower clock limit, to make their tool more useful. They said, it wouldn't be in the spirit of Afterburner.. MrS I'm currently running Ray Adams ATI Tray Tools to control the ATI GPU clocks and Afterburner to control NV and the fans. It works but would be nice to have only 1 tool. |
![]() Send message Joined: 1 Sep 08 Posts: 204 Credit: 219,354,537 RAC: 0 ![]() ![]() |
I'm also using ATI Tray Tool because it's about the only tool which lets me downclock as far as I want. And I'm usually driving my display from the Intel GPU integrated into the CPU (no screen lag), unfortunately ATI Tray Tools doesn't detect my AMD at all this way. Are there any other tool which can do both, downclock (~50%) and detect / set all GPUs? MrS Scanning for our furry friends since Jan 2002 |
![]() Send message Joined: 2 Sep 07 Posts: 24 Credit: 26,244,596 RAC: 0 ![]() ![]() |
TRIXX allows me to select a memory speed on my 7970 between 600 and 2000 |
Send message Joined: 17 Nov 10 Posts: 12 Credit: 482,404,183 RAC: 1 ![]() ![]() |
I'm getting 55 sec. runs on my 6950's.Kurt Sorry abou that,I meant 65 sec.(fat fingers) |
![]() ![]() Send message Joined: 15 Jul 08 Posts: 383 Credit: 729,293,740 RAC: 0 ![]() ![]() |
I'm also using ATI Tray Tool because it's about the only tool which lets me downclock as far as I want. And I'm usually driving my display from the Intel GPU integrated into the CPU (no screen lag), unfortunately ATI Tray Tools doesn't detect my AMD at all this way. Are there any other tool which can do both, downclock (~50%) and detect / set all GPUs? MrS Not sure what you mean, doesn't detect the AMD GPU cards or an AMD IGP? I disable the IGPs as they're' too slow to be of use and run too hot anyway. I don't use any Intel CPUs at this point (except my 3.5 year old core2 duo notebook that's been running at 100% CPU since I bought it) but there's no lag at all on any of my AMD systems as long as I put --gpu-target-frequency 105 in the cmdline when running 2x MW. None of the other projects need any special parameters and no lag there either. |
Send message Joined: 28 Feb 10 Posts: 120 Credit: 109,840,492 RAC: 0 ![]() ![]() |
But in Afterburner you can't downclock the RAM by much.. as it would be efficient for MW. I even asked them some time ago to somehow let people adjust the lower clock limit, to make their tool more useful. They said, it wouldn't be in the spirit of Afterburner.. Have activated the unofficial overclocking mode? With that I come down on my 4850 till freezing screen. look here http://forums.guru3d.com/showthread.php?t=338906 regardes Franz |
![]() Send message Joined: 24 Sep 11 Posts: 4 Credit: 7,304,504 RAC: 0 ![]() ![]() |
Crunching on my 7970 now. 60s each WU (stock clock). Not that much of a improvement from my 5850 (76s), but considering its a nvidia app, i guess its good. :) |
![]() Send message Joined: 24 Feb 09 Posts: 620 Credit: 100,587,625 RAC: 0 ![]() ![]() ![]() |
Set it for 1125/950 and power slider to max 20% it will take that without breaking sweat. Regards Zy |
Send message Joined: 16 Mar 09 Posts: 11 Credit: 80,957,300 RAC: 0 ![]() ![]() |
Thanks, I've got mine running using the nvidia app as well. Is the reason that you are decreasing the RAM speed to increase reliability? So far I've had 2 invalid from 130, so would decreasing RAM speed decrease process speed by more than about that 1.5%? I've put the GPU clock up to 1125 + TDP up 20% and left RAM on stock. |
![]() Send message Joined: 24 Feb 09 Posts: 620 Credit: 100,587,625 RAC: 0 ![]() ![]() ![]() |
.....Is the reason that you are decreasing the RAM speed to increase reliability? Bit of a long reply, as I wanted to include some whys and wherefores so you understand better whats going on. No, its to use less bandwidth on the Card memory. A GPU is dependent on a CPU, it cannot operate on its own. To do its task data goes back and forward between the CPU and GPU. The amount of data that travels between the two will vary from application to application, and for many reasons. The bottom line is that such data travel doesnt happen often with GPU applications, and when it does, it can be a small ammount of data. In MWs case its very small, so it does not need the bandwidth. Setting the bandwidth high on MW GPU applications is pointless as it achieves nothing (the setting is a measure of bandwidth, not speed), just eats more power. So you understand why and whats going on .... If it was Collatz for example, different story. With 5XXX cards, collatz reserves 30% of the card memory for one GPU application, so a 1Gb card has 300mb reserved, it also reserves the same amount of memory in the main system memory (300mb). It uses the 300mb main memory to collate data before sending it to the GPU, and the 300mb in the Card to receive it. 300mb is a real chunk to be tossing around, even for a PC, so Collatz benefits from high Bandwidth on the card (anything up to circa 1300+mhz). If the memory width doesnt take it all in one go, it splits it into two chunks, therefor slowing the application. MW has no such worries, I think its around 30Kb, not sure, but small enough not to worry about it or need high memory bandwidth. The unknown (as such to us mere mortals) at present, is the behavious of the MW app as an OpenCL app with 7970s - but its looking like the same story, very low memory needs on the card, so no need to set high memory, just eats power. There is a complication though due to card design. AMD have gone to extreme lengths to stop a repetition of the 5970 heating problems overclocking, and there is a pretty heavyweight hardware based control over voltage and memory clock speed, especially voltage. You may find the card will refuse to go below a set memory level, have to live with it, its by design, and no matter what it sometimes shows on third party tool screens, its reset itself to default memory (1375). Its looking like the way point is around 1100/1200 memory, so if you find its resetting to default (1375), tweek it up a little to say 1200, and its fine. Cards that are "Black Edition" dont have that level of issue as they are made for overclocking. So far I've had 2 invalid from 130, so would decreasing RAM speed decrease process speed by more than about that 1.5%? With MW there will be no speed decrease, for the reasons outlined above. Think of it as (say) a six Freeway around the city. When its busy, cars back up the entry ramps trying to get on the Freeway, and when they do its all log jammed, just too much traffic trying to move at the same time - if the Freeway "Bandwidth" was higher (ie 18 lanes not 6, there would be no hassle. In none rush hour terms, maybe only a couple of lanes are busy, the other four are (say) empty. Once you get on the freeway a 44tonne truck will go same speed as a lightweight car (lets not get picky here rofl). Its the same with data going from CPU to GPU - memory setting is about how much data passes a point at anyone time, not the speed it flows. I've put the GPU clock up to 1125 + TDP up 20% and left RAM on stock. That will work fine, once you've settled and feel more comfortable with it, turn down the memory by 50mhz steps, until the memory resets itself to 1375. At that point, go 50mhs up, and thats as low as you will go with a non-black edition card. Be aware the 7970 has Massive untapped speed under the hood, there is still around +60% performance available there. It however will only be unlocked by hardware changes by AMD Partners building non reference cards, or by (to a certain degree) voltage tweeking with black edition cards. Dont go near voltage tweeking until you understand more about it all, its a quick way to melt the card. Playing with the power slider in the driver is absolutely fine, card is safe, just stay away from external voltage tweeking for now. E-Peen is one thing, ruined bank accounts replacing melted cards is another :) Regards Zy |
![]() ![]() Send message Joined: 15 Jul 08 Posts: 383 Credit: 729,293,740 RAC: 0 ![]() ![]() |
Wow, nice answer Zy. Tells us more about the 7970 cards for sure. |
![]() Send message Joined: 24 Feb 09 Posts: 620 Credit: 100,587,625 RAC: 0 ![]() ![]() ![]() |
Hardcop did a good no holds barred session overvolting the 7970 - results are stunning, and serve to illustrate whats under the hood of a 7970, and just how much more power is to come as partner cards come out. The upfront published performance figures of reference cards tell no where near the whole story of 79XX, the hardcop test does. http://hardocp.com/article/2012/01/09/amd_radeon_hd_7970_overclocking_performance_review 2012 gonna be fun as all this unwinds and the green team add their 28nm offering into the fray. Moral of the tale .... take a deep breath, let it all unwind and dont rush into a buy unless of course you have to and need a card now. Waiting for 79XX AMD Partner cards (a month or two probably) and seeing how NVidia 28nm performs end March/start April (if for once NVidia marketing guys are not telling porky pies on deployment dates - an NVidia habit) is a very wise choice to make. Dont let e-peen take over, there is much to come yet in the 28nm world, and we've not even started on the new AMD GCN architecture, or the changes the green team make to their architecture. The move from 40nm to 28nm cards open up a whole new GPU world made even more complex by the decision by AMD and NVidia to standardise on OpenCL, dont (as such) measure it by 40nm 5XXX standards, you'll trip over. Patience on this will be rewarded, rush buys will just empty bank accounts :) Regards Zy |
![]() Send message Joined: 1 Sep 08 Posts: 204 Credit: 219,354,537 RAC: 0 ![]() ![]() |
Thanks for answering, guys. - Trixx seems fine for clocking the memory down - yes, I had enabled the unofficial OC in Afterburner.. but never mind Not sure what you mean, doesn't detect the AMD GPU cards or an AMD IGP? I disable the IGPs as they're' too slow to be of use and run too hot anyway. What I'm doing: I've got the DVI input of my monitor hooked up to the Sandy Bridge GPU and running all normal stuff, videos, light games etc. The ATI is connected to the VGA port (don't have a display port cable..), so windows keeps it active, and is crunching. I'm only disturbing it for demanding games.. occasionally. The problem: If I want to play a game on the ATI all I'd have to do is stop BOINC GPU and start it, Virtu will push the output to the IGP. However, I also have to adjust the clocks: core down, memory up. And this is what's not working: in my standard config (IGP @ DVI as main display) I can't set the clocks on the ATI. ATI Tray Tools doesn't find the card (says no ATI present), the MSI Tool crashes, Trixx seems to work but doesn't actually change anything. What I do now is switching the primary desktop, switching to the VGA input on my screen, changing the clocks, replacing the primary desktop and switching back to DVI. Rather cumbersome.. Any further ideas? Best regards, MrS Scanning for our furry friends since Jan 2002 |
Send message Joined: 16 Mar 09 Posts: 11 Credit: 80,957,300 RAC: 0 ![]() ![]() |
Thanks loads for the comprehensive explanation Zy, that's appreciated. I've got a much clear understanding now. I'd read the black edition cards were coming out pretty soon (saw something about one with a water block on it will be set at 1.3 out of the box!), but had been waiting so long for the 28nm transition I was foaming at the mouth. Sadly having a limited budget and also limited space (only a few pc's and they have only so many slots) - I'm holding out for a 7990 (new zealand) due in about march |
![]() ![]() Send message Joined: 15 Jul 08 Posts: 383 Credit: 729,293,740 RAC: 0 ![]() ![]() |
What I'm doing: What I'd do: disable the SB IGP. Don't see what you're gaining by using it except more heat. Other than that.. |
![]() Send message Joined: 24 Feb 09 Posts: 620 Credit: 100,587,625 RAC: 0 ![]() ![]() ![]() |
..... I'm holding out for a 7990 (new zealand) due in about march Thats what I've got earmarked for my main box, 7990s, however dont discount NVidia, they have come up short with the 3XX & 4XX debacles over a four year period, but managed a last minute rescue with 560/570/580. They are capable of coming out with competitive hardware, lets wait for reality end March/start April. The big question in BOINC terms is what the software side of life will bring, and at present thats a big unknown in BOINC terms. Need to be careful dont get too limited on BOINC Projects during this early stage of transition to OpenCL, and that transition will last for a while, its not all going to happen at once. Could take a year or more to pan out ..... OpenCL ...... CUDA ...... Stream ...... GCN ....New NVidia architecture ..... which way will it all turn and when ..... so, decisions decisions rofl :) Regards Zy |
Send message Joined: 16 Mar 09 Posts: 11 Credit: 80,957,300 RAC: 0 ![]() ![]() |
Yeh just read early prediction the new kepler architecture is rumoured to be faster than GCN (based in hands on tests) by guy at semiaccurate who is usually vv pro amd. It could be time for me to stop thinking AMD gpu's only (I had a few nvidia ones fail and while back and got my knickers in a knot about them for a few years). The anticipation! OooOoooo Just loaded some drivers for the 7970 and it's dropped the process time from 62 to 53s (running +20 thermal envelope and 1125MHz). Under ideal scenarios they can improve OpenCL by 250%. Link is: http://support.amd.com/us/kbarticles/Pages/RC11Driver.aspx |
©2024 Astroinformatics Group