Milkyway@Home - GPU & CPU performance stats wanted, any capable h/w, old or new!

[Assimilator1]

[update 4] Well maybe not then, the long WUs have made some what of a comeback! But who knows how long for?......

[update 3] This benchmarking run is now over.
What started off as a singular speed 213.76 credit WU became 2 different speeds by late February, our 'standard' WU was joined by a much quicker one (approx. 1/3 faster). Our standard 'long' 213.76 credit WU became rare 2 months latter & is now extinct, hence the end .

***************************************************************************************************

Following on from the original thread, I thought it was time for an update & to get some new MW GPU scores .

Incase your wondering why the need, well Sunny129 mentioned somewhere that WUs times have changed radically in the nearly 3 yrs since that thread was last updated & so unfortunately it is obsolete, (although you can still compare those GPUs relative performance).

So I did a quick comparison, sticking to the 213.76 credit WU as a benchmark. In the old thread an HD 5850 was doing them in ~115s whereas my HD 5850 is doing them in ~295s, even allowing for the fact that my GPU is only at ~80% load (it doesn't have a dedicated CPU core atm) that's still a huge increase in crunching time! (I'm going to dedicate a core for it btw for a while to give a proper benchmark score).

Oh & I'll be doing this table in seconds, much easier to calculate/compare & afterall the online tasks page shows the times in seconds!

So please share your new scores for old & new GPUs & CPUs alike!
I will update the table as necessary.

Requirements for benchmark, validated 213.76 credit WUs only, average of 5 WU times, dedicated CPU core for the GPU, please state clock speeds if overclocked (including factory o/cs) or state 'stock'. It would also be handy if you could state your BOINC & driver version & OS, incase it does make any odds.

[update] It seems MW@H have now (20th Feb. ish) released new faster 213.76 credit WUs along with the previously 'standard' speed WUs, so going by the benchmarks gathered upto the 17/2/14, you need to average 5 of the longer 213.76 credit WUs only. Each 'speed' WU should only vary by a few seconds or so where their is nothing else working the GPU.

[update2] As of 21/4/14 the long 213.76 credit WUs are extremely rare now so dig deep to find them! The 2 times I got today I found 5 in 307 & 5 out 1020 results!! Once these WUs are gone it'll be the end of this benchmarking run.

Oh & for the benchmark please only crunch 1 WU at a time per GPU otherwise it will massively increase WU time, naturally! (even if it does increase output), and the WU times seem to fluctuate much more than singly crunched WU so you can't 1/2 the time either. For CPUs you'll want to crunch 1WU per real core.

Current GPU statistics ~ Average Run Time to Complete 1 'long' 213.76 credit WU :-

HD 7970 (GPU 1150 MHz) ....................................... 75s Sunny129
R9 280X (GPU 1100 MHz) ........................................ 75s Mumak
R9 280X (GPU 1030 MHz) ........................................ 77s Dunx
HD 7970 (GPU 1100 MHz) ....................................... 77s Sunny129
HD 7970 (GPU 1000 MHz, RAM 1400 MHz, Cat 14.1) .. 78s Mesyn191
HD 7950 (GPU 1100 MHz) ....................................... 80s Leonheart
HD 7950 (GPU 1100 MHz) ....................................... 81s Mumak
HD 7970 (GPU 1050 MHz) ....................................... 81s Sunny129
HD 7950 (GPU 900 MHz) ......................................... 82s WES
HD 7970 (stock) .................................................... 83s mikey
HD 7970 (GPU 1000 MHz) ....................................... 85s Sunny129
HD 7950 (GPU 1000 MHz, RAM 1500 MHz,Cat 14.1 b) . 87s Leonheart
HD 7970 (GPU 950 MHz) ......................................... 90s Sunny129 (same time when RAM u/c to 800 MHz!)
HD 7950 (GPU 1000 MHz, RAM 1500 MHz) ................ 95s Leonheart
HD 7870 XT (stock) .............................................. 120s Matt
HD 7950 (GPU 850 MHz) ....................................... 121s salvordorhardin
HD 6970 GPU 950 MHz, RAM 900 MHz) ................... 148s Brickhead
HD 6990 (stock) .................................................. 165s Alez [TSBT's Pirate] (dual GPU card)
HD 6970 (stock) .................................................. 165s Mikey
HD 6950 (stock) .................................................. 176s Icecold
HD 5870 (GPU 868 MHz, RAM u/c 900 MHz) ............ 187s Brickhead
HD 5870 (stock) .................................................. 192s mikey
HD 5850 (GPU 850 MHz) ....................................... 210s Assim1
HD 5850 (GPU 850 MHz, RAM u/c 500 MHz) ............ 211s Assim1
HD 5850 (GPU 775 MHz) ...................................... 231s Assim1
HD 5850 (GPU 750 MHz) ...................................... 245s salvordorhardin
HD 5850 (stock) ................................................. 246s Assim1 (this post)
HD 5850 (GPU 800 MHz) ...................................... 246s petrusbroder
HD 5830 (stock) ................................................. 282s Pheonix
GTX 780 Ti (stock) .............................................. 403s ... Linux - 395s biodoc
R9 270X ............................................................ 407s Alez [TSBT's Pirate]
GTX Titan .......................................................... 414s Yankton (GPU load only 18% despite a free CPU core)
HD 4870 512 MB (stock) ...................................... 444s JumpinJohnny
GTX 560 Ti 448c (GPU 880 MHz) ........................... 471s GleeM
HD 4870 1GB (stock) ........................................... 503s Assim1
GTX 570 (stock) .................................................. 520s - Linux biodoc
GTX 480 (GPU 750 MHz) ...................................... 520s Dunx
HD 7970M .......................................................... 525s Spencer
HD 4850 1 GB (stock) .......................................... 553s wayliff
GTX 770 (GPU 1333 MHz) .................................... 609s Stojag
HD 4830 512 MB (GPU 670 MHz) .......................... 615s Assim1 (this post)
HD 7770 GHz ed. (GPU 1100 MHz) ........................ 724s Mumak
HD 7770 GHz ed. (stock) ..................................... 807s Deerslayer
GTX 660 Ti (GPU 1046 MHz) ................................. 816s - Linux biodoc
GTX 560 Ti (stock) .............................................. 836s Deerslayer
HD 7750 (GPU 900 MHz, RAM 1300 MHz) ............. 1096s branjo
GTX 460 (GPU 750 MHz) 768 MB ......................... 1127s - Linux Ken g6

Current CPU statistics ~ Average Run Time* to Complete 1 'long' 213.76 credit WU :-

Intel Xeon E3-1230 v3 (Turboing to 3.7 GHz) ...... 9881s Stojag
Intel Core 2 Q9550 (3.6 GHz, 424 MHz FSB) ..... 12,532s Assim1
Intel Core i7-3612QM ..................................... 13,068s mikey
AMD Phenom II X4 965 (3.41 GHz) .................. 13,774s JumpinJohnny
Intel Core 2 Q8300 ........................................ 14,949s mikey
AMD Phenom II X6 1090T ............................... 14,950s mikey
AMD FX 6300 ................................................ 15,521s mikey
AMD FX 6100 ................................................ 17,449s mikey
Intel Core 2 Q8200 ........................................ 19,542s mikey
AMD Phenom 9850 ........................................ 25,844s mikey

OS Windows, unless otherwise stated. CPU at stock clock where no speed shown.

* I latter realised for the CPU tests only that I should of used & asked for the 'CPU time' rather than 'run time' which would include any slow downs due to running other processes. I think that largely due to the use of averaging at least 5 WU times that it will keep the inaccuracy low & that the times are still useful, just bear in mind their will be a little inaccuracy there. Just to re-iterate that this refers to the CPU benchmarks only.

I've also created a thread at the MW forums http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=3465 , although the daft 1hr edit limit might means I can only post updated tables at the end of the thread!

And a new thread at the KWSN forum
And at OcUK forums

******************************************************************************

I'll kick things off with my sons rig :-
HD 4830 (which does have a dedicated core), cat 12.6, GPU 670 MHz, RAM stock 900 MHz, Win 7 ult. 64bit. - 615s

[update] My HD 5850 1GB, stock, Cat 13.1, Win 7 Ult. BOINC 7.2.33 - 246s

 

[deerslayer]

Currently I run Rosetta on my cpu, and MW@H gpu only. When I get some time I'll switch settings and suspend Rosetta and get you some info.

Edit: the most recent 213.76 wu's I saw on your 5850 rig were around 320 sec. Am I looking at the wrong thing?

Also, when my wus are running they say 0.9 cpu. The rig that is running these has an old dual core processor, and both cores are running Rosetta, so now I'm a bit confused (not unusual).

 

[Assimilator1]

You only need to dedicate 1 CPU core per GPU, so for that pc you'd just reduce R@H CPU usage to 50% in BOINC preferences.

I've got all 4cores running A@H & LHC@H again now, hence the MW times going back up.......... more than at 1st which I don't understand . The 246s time was with a dedicated CPU core.

No idea what that 0.9 cpu means in BOINC , any1 else?

 

[deerslayer]

Well, if I can ever get my video card drivers right in Ubuntu I'll test it out. Not having any luck so far. BOINC can't detect any useable gpu.

 

[Ken g6]

Average of 10 WUs on a GTX 460 768mb OC@750: 410.937s

This seems like it's a lot faster than it used to be. BOINC Manager seemed to think the WUs would take about half an hour when it started. Did they improve the client for nVidia?

 

[Assimilator1]

Thx for the result ken , good question re Nvidia, IDK TBH, if no one else (Sunny?) has replied in the next day or 2 I'll have a dig around the MW forums etc & try to find out.

Btw it is the GPU that's o/c to 750 right? You didn't specify .

 

[deerslayer]

You only need to dedicate 1 CPU core per GPU, so for that pc you'd just reduce R@H CPU usage to 50% in BOINC preferences.

I've got all 4cores running A@H & LHC@H again now, hence the MW times going back up.......... more than at 1st which I don't understand . The 246s time was with a dedicated CPU core.

No idea what that 0.9 cpu means in BOINC , any1 else?

How would I do this on my FX-6300? Is this my Rosetta preferences or preferences somewhere else?

 

[Assimilator1]

I've no idea how many cores an FX has, but you just divide the 100% BOINC CPU usage by the number of cores you have, then reduce the 100% to leave 1 core free.

In BOINC client (advanced view) >tools>computing preferences, 1st tab is processor usage, find field 'On multi processor systems, use at most % of ...'
Then set % to leave a core free, so it would be 75% for a quad core. If you were to look in task manager after doing that you'll probably see CPU usage barely go over 75%, but except for perhaps older GPUs (e.g HD 4800s) or slower GPUs it'll be allowing the GPU to crunch at it's fastest (confirmed by looking at GPU load under Cat (CCC) drivers performance>AMD overdrive). GPU load should be 98%+ ish.

I found my 4870 was at near 100% GPU load without a dedicated CPU core running F@H on the CPU, but my 5850 is only around 80% loaded like that. Though strangely my sons PC running Asteroids & MW it varied between 40-98%! . So it's best just to dedicate a core for benchmarking at least.

 

[salvorhardin]

I averaged 121s on a 7950 @850/1250 on catalyst 13.9, windows7, boinc 7.2.38 and with 2 cores free on a stock i7 3770k. My 5850 #750/1000 averaged 245s on catalyst 13.9, windows7, boinc 7.0.33 and with 1 core free on a stock q9400.

 

[deerslayer]

Purely for reference, before I changed my processor usage, my 7770 had done 48 of the 213.76 credit wus @ an average of 984.2 seconds.

Boinc 7.2.33, Windows 7 Pro 64, HIS Radeon HD 7770 ghz edition. (stock 1000 core / 1125 memory - 1gb 128 bit gddr5) Will report back when I get some results to check with the cpu settings changed.

 

[Assimilator1]

The 984s time, I take it that's with no dedicated CPU core?
Be interesting to see how much difference it makes to your GPU.

Btw I just found out that Asteroids@H was causing my GPU load to drop to 40-75%, so I've had to dedicate a core to MW permanently now, gain for MW but a loss for A@H.

salvorhardin
Interesting to know you can run your 5850 on Cat 13.9, I'll have to try that 1 sometime (did I already say that in another thread? lol).
And thx for your results

 

[salvorhardin]

I was having the same problems other people were having with mw wus crashing and tried updating the drivers. Doing that screwed up the computer so that any driver I install causes my computer not to boot into windows I was able to do a system restore but 13.9 is the only driver that boots windows and shows up as opencl in boinc. I usually notice a 50% increase in runtimes when I don't dedicate a core to the gpu but since it uses less of the gpu there is less windows lag.

 

[wayliff]

Windows 7 Pro SP1 x64
Boinc 7.2.28 x64
AMD Catalyst 3.0.868.0 - Installed on March 23, 2012.

HD4850 1024MB RAM
GPU Clock @ 625Mhz and Memory @ 993Mhz (These are stock values)

Average of 5 WUs is 553.49s

 

[Assimilator1]

Thx wayliff :thumbsup:

 

[deerslayer]

HIS Radeon HD 7770 ghz edition = 806.6 sec
Windows 7 Pro 64 bit
Boinc 7.2.33
13.251 amd driver

Gigabyte GTX 560 ti 900mhz = 836.4 sec
Windows 7 Home Premium 64 bit
Boinc 7.2.33
331.82 nvidia driver

And I discovered the hard way that a 6850 is not compatible with MW@H. Any suggestions on something I CAN run on it?

 

[Ken g6]

Gigabyte GTX 560 ti 900mhz = 836.4 sec
Windows 7 Home Premium 65 bit
Boinc 7.2.33
331.82 nvidia driver

I think something's not right there. Maybe if you went back to an earlier driver it would be faster?

I'm an odd duck: I'm running the Linux 325.15 driver, which I don't think was ever officially released. But it works for me.

And I discovered the hard way that a 6850 is not compatible with MW@H. Any suggestions on something I CAN run on it?

PrimeGrid's PPS Sieve will run on anything that runs OpenCL, back to the 4000 series. But there ought to be something better.

 

[deerslayer]

I wasn't really impressed with the times on either one of them. Apparently my fat fingers have prevailed again. I have an extra bit in my windows.

 

[GLeeM]

I think something's not right there.

I think he meant your time, not the typo.

Compared to his it looks like something is wrong? Your newer card should do better than his older card? Or maybe his number is off?

On my GTX560 Ti 448 core at 880Mhz = 470.7 seconds
Win7 64 bit, BOINC 7.0.64, driver 306.97

I'll try again to get some of those WUs on my 7950.

 

[deerslayer]

I think he meant your time, not the typo.

Compared to his it looks like something is wrong? Your newer card should do better than his older card? Or maybe his number is off?

On my GTX560 Ti 448 core at 880Mhz = 470.7 seconds
Win7 64 bit, BOINC 7.0.64, driver 306.97

I'll try again to get some of those WUs on my 7950.

I know. When I saw the time I didn't think it was right either. I was just commenting on my epic 65 bit windows installation.

 

[Ken g6]

Or maybe his number is off?

See for yourself:

Computer
Tasks

And a lot of the later tasks were without a dedicated core.

 

[Assimilator1]

How have you managed to get only 159.86 credit WUs? lol

Gleem & Deerlsayer, thx for your results.

Deerslayer
I'll hold off on your 560 Ti time for now, as the other guys mentioned that time does seem to be off. Maybe driver issue or GPU utilisation issue??
I know on F@H my GTX 260 c216 (& others) was better off with 323 drivers (IIRC!).

 

[deerslayer]

I'll try a different driver when I get ambitious and report back.

 

[Assimilator1]

Oh & lol at 65 bit windows

 

[deerslayer]

Oh & lol at 65 bit windows

You guys are just jealous.

 

[Assimilator1]