Welcome to MilkyWay@home

Posts by ExtraTerrestrial Apes

41) Message boards : Number crunching : 79XX Dont Run (Message 52655)
Posted 26 Jan 2012 by ExtraTerrestrial Apes
Post:
May have to reluctantly go to Win 7 or just forget about Upgrading to the 7970's ... :/


If you have to buy several licenses that'd be painful. If you can *try* one for free, e.g. over MSDNAA, that might be a less painful start. If you look at it open minded (e.g. you do not want everything to be exactly as it has always been) it's actually amazing how much things have improved over the years.

I used to hate how MS always tried to think for me and do everything automatically.. and never got it quite right. With Win 7 they're finally getting there, I can lay back and mostly just let it do its thing, without me having to worry about stuff. I tried it when it came out and never looked back.

If it's just for a cruncher you could consider getting used Vista licenses. The graphics driver model is the same as for Win 7, so will last a lot longer than XP. For actual work it's much less pleasant than Win 7, though.

MrS
42) Message boards : Number crunching : 79XX Dont Run (Message 52623)
Posted 23 Jan 2012 by ExtraTerrestrial Apes
Post:
@far: your card is hot and loud already. I would't want to push it any further, even though the chip has reserves. Improved case cooling may help, or maybe replacing the "thermal interface material" on the chip.. although this is somewhat dangerous and is usually only worth it if something's wrong with the factory applied stuff.

I your situation I'd probably give Afterburner a shot and try to lower the GPU voltage. Massive OC headroom means you can also stick to the frequency and lower the voltage a lot -> significantly reduced power consumption and heat generation. You could also go for a soft OC at slightly lower voltage (better performance, same power -> more efficient).

@pressure fans: they're trading maximum air flow for higher static pressure. Fans can only achieve their rated air flow if nothing impedes the flow. Obstackles could be tight CPU cooler fins, dust filters or lot's of cables. In such situations the normal fan might only be able to create 30% of its rated air flow, whereas a pressure fan might be able to realized 50% of its maximum. Whether you want pressure fans as case fans really depends on your case geometry.
BTW: interesting read I had about this yesterday.. never heard of it before! Way better than working on.. other stuff :p

MrS
43) Message boards : Number crunching : GPU reference - Time to complete a WU (Message 52600)
Posted 22 Jan 2012 by ExtraTerrestrial Apes
Post:
For what its worth, I think the current sweet spot is the Gigabyte 6950 1GB: $240USD for a theoretical 2200 GFlops (single precision - double is around 500) I run them at 900MHz clock and +20% power and they generally do MW tasks in about 63 seconds.


An even sweeter spot is an HD6950 2GB which unlocks to the full HD6970 shaders. Mine is running at 900 MHz too (no + power tune, it doesn't even com close to 250 W anyway) and crunching single 160 credits WUs in 53 s :)

That's a bit shy of 260k RAC at unperturbed 100% MW.

MrS
44) Message boards : Number crunching : 79XX Dont Run (Message 52594)
Posted 21 Jan 2012 by ExtraTerrestrial Apes
Post:
- Folding@Home: intermittent stability issues have been resolved for the OpenCL version of Folding@Home


That means Folding runs on GNC now? Anyone got any numbers?

MrS
45) Message boards : Number crunching : 79XX Dont Run (Message 52569)
Posted 18 Jan 2012 by ExtraTerrestrial Apes
Post:
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
46) Message boards : Number crunching : Help finalizing video card selection (Message 52529)
Posted 15 Jan 2012 by ExtraTerrestrial Apes
Post:
Let's call it an "order of GPU performance" then :D

MrS
47) Message boards : Number crunching : 79XX Dont Run (Message 52528)
Posted 15 Jan 2012 by ExtraTerrestrial Apes
Post:
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
48) Message boards : Number crunching : Help finalizing video card selection (Message 52521)
Posted 15 Jan 2012 by ExtraTerrestrial Apes
Post:
They are 28nm cards and therefore far far more power efficient than previous series by an order of magnitude in comparison to what has been normal in cards so far.


That would be a factor of 10. A bit generous ;)
Might be that high if GCN is a huge improvement over VLIW for some app (probably GPUGrid), but not for something which runs well on VLIW (e.g. MW, Collatz, Moo).

MrS
49) Message boards : Number crunching : 79XX Dont Run (Message 52520)
Posted 15 Jan 2012 by ExtraTerrestrial Apes
Post:
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
50) Message boards : Number crunching : HD7970 on the horizon .. (Message 52493)
Posted 14 Jan 2012 by ExtraTerrestrial Apes
Post:
If they run nbody simulations, fine. Otherwise it's an even greater waste of valuable resources (IMO).

MrS
51) Message boards : Number crunching : HD7970 on the horizon .. (Message 52481)
Posted 14 Jan 2012 by ExtraTerrestrial Apes
Post:
Isn't those times faster than any of the NVidia cards?


Forget nVidia for MW, it's a waste of compute resources.

MrS
52) Message boards : Number crunching : Help finalizing video card selection (Message 52480)
Posted 14 Jan 2012 by ExtraTerrestrial Apes
Post:
Don't worry about the PSU for your sons machine. Unless the PSU is complete crap, it will handle it just fine. And if it's crap - you probably wouldn't want to run it anyway.

Regarding the card: buying a new budget card for MW seems like a bad idea to me. They're so weak, they're hardly worth it at all. Even if the nVidias weren't crippled at double precision. If the choice was between GT520 and GT430, definitely go for the 430. It's noticeably more powerful. However, I wouldn't run MW on any of them (there are other nice CUDA projects which work with single precision).

If I were you I'd probably look for a more powerful, used GPU. Recently I sold me HD4870 (was doing a little shy of 100k RAC at MW) for 30€ (friend price), as it isn't power efficient any more for me to run it 24/7. Now it's in a budget gaming box and should serve this role well for some time to come.

Regarding your other update plan: Mikey's 585 W PSU was probably completely crap. It's not normal for a system with a mere HD5770 (~100 W at full load.. which it is not drawing upon system boot) to require a 650 W PSU.

Personally I've got an HD6950 with unlocked shaders @ 900 MHz (no voltage increase), crunching MW 24/7. The CPU is an i7 2600K @ 4.0 GHz (moderate voltage). The PSU is an Enermax 80+ Gold 500 W. I'm drawing < 300 W from the wall. It's exhausting a soft breeze of cold air - easy.

Adding a 2nd GPU would push power draw close to 500 W (at the wall, so ~455 W from the PSU), uncomfortably close. A quality 650 W unit should be able to handle it, though. 250 W is the maximum allowed GPU power draw, not the actual one. You might feel more secure with a higher rated one, but 1 kW is definitely overkill for 2 GPUs.

Wouldn't your HD4670 be the perfect card for your son?

You might also want to consider the new AMDs with Graphics Core Next architecture. They're brand new, so we can't say how good they are at crunching yet. But that's for sure: they're way more flexible than the current lineup. They will be much more useful down the road.

MrS
53) Message boards : Number crunching : HD7970 on the horizon .. (Message 52346)
Posted 9 Jan 2012 by ExtraTerrestrial Apes
Post:
Anyone got one? I wonder if the current app will have to be adjusted, or if it just works.

MrS
54) Message boards : Number crunching : HD5870 dead - help with replacement! (Message 50890)
Posted 30 Aug 2011 by ExtraTerrestrial Apes
Post:
Imagine it was a car, or mobile phone, or fridge, would or could anyone wait over 10 weeks for repairs???

OT: a colleague of mine is waiting since 1 year for his new car. It's some Kia, so maybe not the best idea in the first place anyway..

MrS
55) Message boards : Number crunching : HD5870 dead - help with replacement! (Message 50854)
Posted 28 Aug 2011 by ExtraTerrestrial Apes
Post:
THe GPU needs assistance every few ms, whereas in NUMA we're talking about a few 10 ns difference in access times to non-local memory. Even in the slow case that's still 5 orders of magnitude faster than what the GPU needs.

So I'm still puzzled. The only possible reason for this performance difference that I can think of is the following: in the old app Gipsel estimated when the GPU might need the next CPU intervention. This was pretty accurate, but he also included an option to scale the calculated wait times by some factor. If the CPU needs much longer for its calculations than expected, the calculated time point when the cpu thread is being woken up might be too late, leading to a GPU starving for data. In this case a simple "scale wait times by 0.95" should do the trick.

Or something else: is MS using a coarser granualarity for the scheduler in multiprocessor systems with NUMA enabled? LEt's say 5 ms instead of 1 ms (standard)? In that case I could see how performance would drop, as the CPU thread might get activated too late.

MrS
56) Message boards : Number crunching : HD5870 dead - help with replacement! (Message 50841)
Posted 26 Aug 2011 by ExtraTerrestrial Apes
Post:
Wow, that's highly unexpected! NUMA stands for "non-uniform memory access", meaning that one Xeon can access the memory of the other one via the quick path link, right? That increases average memory latency and improves overall bandwidth. I don't see how this could slow a GPU down that much. Maybe a bug in the BIOS? Is anything else running faster now?

MrS
57) Message boards : Number crunching : HD5870 dead - help with replacement! (Message 50827)
Posted 24 Aug 2011 by ExtraTerrestrial Apes
Post:
Wow, 2 6-core Xeons for 24 threads overall. That's a lot of thread juggling. Could it be that your MW CPU process is starved for memory bandwidth? As Len already suggested, I'd vary BOINC CPU usage to see until when I'd get good GPU usage (using interval halfing, i.e. start with 0 cores, then 12 and then 6 or 18 depending on whether 12 is still good).

MrS
58) Message boards : Number crunching : HD5870 dead - help with replacement! (Message 50813)
Posted 23 Aug 2011 by ExtraTerrestrial Apes
Post:
Several people have claimed on here that leaving a whole core available for MW ATI app results in improved throughput. I'm not seeing it...

Absolutely. With all cores loaded I'm either seeing full utiliztion or ~75% on my HD6950 with unlocked shaders in Win 7 64. It somehow depends on the CPU projects, but I haven't been able to identfy which one is to blame.

If I leave one logical core of my i7 free I'm getting consistent 98-99% utilization. I'm using
app_info.xml wrote:
<cmdline>--process-priority 3 --gpu-disable-checkpointing</cmdline>

but I don't think it matter much here. Feel free to check out the performance of that card - that's totally impossible for a Cypress (without liquid nitrogen cooling). What you're seeing is some weird software problem not restricted to Cayman. I suppose you'd see similar figures, just the other way around, if you switched the cards between the PCs.

MrS
59) Message boards : Number crunching : Advice request (Message 50740)
Posted 18 Aug 2011 by ExtraTerrestrial Apes
Post:
For crunching any NV cards smaller than the GTX560 don't make much sense, if your system can take it (power supply, case cooling, space). I wouldn't bother running Mw on it, though. The ATIs are much more efficient at MW, while they can't run the other projects.

MrS
60) Message boards : Number crunching : Need help to use less cores (Message 50715)
Posted 16 Aug 2011 by ExtraTerrestrial Apes
Post:
Your GPU could do more-or-less worthwhile work at Collatz, PrimeGrid, Folding@Home (non-BOINC) or DNETC(*).

MrS

(*) Which I think is completely useless.


Previous 20 · Next 20

©2024 Astroinformatics Group