MilkyWay@H Benchmark thread Spring 2014 - Summer 2016

[Assimilator1]

Benchmark thread Spring 2014 - Summer 2016 (can't seem to edit the title)
[update]
At some point through the summer of 2016 the MW v1.02 app & it's 106.88 credit WU were given the chop, so this benchmarking run is over. I will probably start a new one with the v1.4x app & its 133.66 credit WUs. New benchmarking thread here.

So please share your new scores for old & new GPUs & CPUs alike!

Note new requirements for the benchmark :-

Please use validated 106.88 credit WU results only, they must be from the MilkyWay@Home v1.0x app (the separation mod fit ones are slower. ~20s on my HD 5850 @850 MHz).

Average of at least 5 WU times (not cherry picked please! ).

A dedicated CPU core for each GPU.

Please state clock speeds if overclocked (including factory overclocks) or state 'stock'.

Please only crunch 1 WU at a time per GPU, otherwise it will massively increase WU time! (even if it does increase output, the WU times seem to fluctuate much more than singly crunched WUs so you can't 1/2 the time either).
I've decided to relent a bit on this, but only for the GTX Titan as it can't achieve anywhere near full load with just 1 WU, I will add a proviso stating this by each Titan's score (which will be derived from the total time crunched, divided by the number of WUs being crunched simultaneously. 8 WUs at once seems to be the choice so far).

For CPUs please state whether Hyper Threading (or equivalents) are enabled or not, times for both states welcomed .

It would also be useful if you could state your BOINC & driver version & OS, incase it does make any difference.

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

Current GPU statistics ~ Average Run Time to Complete 1 MW v1.02 106.88 credit WU :-

HD 7970 (GPU 1250 MHz, RAM 1550 MHz) .............. 21.02s .. tictoc
HD 7970 (GPU 1200 MHz, RAM 1550 MHz) .............. 21.3s ... tictoc (Win 7)
R9 280X (GPU 1185 MHz, RAM 1600 MHz) .............. 23s ...... DutchDK
R9 280X (GPU 1125 MHz, RAM 975 MHz) ................ 23.1s ... Arivald Ha'gel
R9 280X (GPU 1100 MHz, RAM 1500 MHz) .............. 23.12s . TennesseeTony (Cat 15.7)
R9 280X (GPU 1180 MHz, RAM 1550 MHz) .............. 23.13s . khrylxtko
GTX Titan Black (GPU 990 MHz) .............................. *23.4s .. Todderbert *Time derived from running 8 WUs at once ÷8
R9 280X (GPU 1100 MHz, RAM 1500 MHz) .............. 23.5s ... TennesseeTony (Cat 14.9)
HD 7970 (GPU 1200 MHz, 1550 MHz) ....................... 23.7s ... tictoc (Win 10, also a different PC)
R9 280X (GPU 1100 MHz, RAM 950 MHz) ................ 24.1s ... Arivald Ha'gel
HD 7950 (GPU 1188 MHz, RAM 1250 MHz) .............. 24.1s ... GLeeM
HD 7970 (GPU 1052 MHz, RAM 1375 MHz) .............. 24.12s .. blub
HD 7970 GE (GPU 1000 MHz, RAM u/c 1375 MHz) .. 25s ...... DrPop
R9 280X (GPU 1070 MHz, RAM 1600 MHz) .............. 25s ...... WES
GTX Titan Black (stock) ............................................. *25.2s ... Sebastian *Time derived from running 8 WUs at once ÷8
HD 7950 (GPU 1080 MHz, RAM 1250 MHz) .............. 27s ...... DR4GOON
HD 7950 (GPU 1000 MHz, RAM 1250 MHz) .............. 28.1s ... GLeeM
HD 7950 (GPU 1050 MHz, RAM 1250 MHz) .............. 29s ..... DR4GOON
HD 7950 (GPU 925 MHz, RAM 1250 MHz) ................ 31.2s ... salvorhardin
HD 7950 Boost (GPU 900 MHz, RAM 1400 MHz) ...... 31.2s ... Kwartet!
HD 7950 (GPU 900 MHz) ........................................... 33s ..... Tom*
HD 6950 (bios unlocked to 6970, GPU 1000 MHz) .... 34s ..... khryl
HD 7870 XT Boost (GPU 975 MHz, RAM 1500 MHz). 34.1s ... kozicka
HD 7870 XT* 3GB (GPU 925 MHz, RAM 1250 MHz) . 37.5s ... Assimilator1 (*sort of an 7870 XT! See here)
HD 7950 (stock) .......................................................... 39.8s ... Rudy Toody
R9 390 (GPU 1015 MHz, RAM 1500 MHz) ................ 41.5s ... salvorhardin
HD 6950 (bios unlocked to 6970, GPU 820 MHz) ...... 42s ...... khryl
HD 7950 (stock) .......................................................... 42.8s ... wayliff
HD 5970 (GPU 900 MHz, RAM 1200 MHz) ................ 43.7s ... superpower (dual GPU card)
HD 5870 (GPU 850 MHz, RAM u/c 600 MHz) ............ 46.2s ... swiftmallard
HD 5850 (GPU 850 MHz, RAM u/c 500 MHz) ............ 51s ..... Assimilator1
HD 5970 (GPU 735 MHz, RAM 1010) ........................ 52.6s ... superpower (dual GPU card)
HD 5850 (stock) .......................................................... 60s ..... Assimilator1
Tesla M2090 (stock) ................................................... 84s ..... fractal
R9 380 (GPU 1010 MHz, RAM 1450 MHz) ................ 96.7s ... andre
GTX 580 (GPU 815 MHz, RAM 1025 MHz) ............... 114s ..... Sebastian
HD 4850 (stock) ......................................................... 121s .... TennesseeTony
GTX 980 (GPU 1417 MHz, RAM 6000 MHz) ............. 130s .... Biodoc (Linux)
HD 7850 (GPU 1030 MHz, RAM 1200 MHz) ............. 139s ... usao
GTX 980 (stock) ......................................................... 145s ... Cookiemonsta (time in post wrongly labelled as 'Calculation of CPU')
GTX 780 Ti (GPU 1084-1111MHz, RAM 6800 MHz) . 157s ... DR4GOON
HD 4830 (stock) .......................................................... 161s .... TLS
GTX 970 (GPU 1114-1253 MHz) ................................ 165s .... Orange Kid
GTX 560 Ti (stock) ...................................................... 252s .... zzuupp
R9 m275 (stock) .......................................................... 290s .... waffleironhead
GTX 460 (GPU 763 MHz, RAM 950 MHz) .................. 292s .... Sebastian
GTX 680 (GPU Boost 1070-1080 MHz) ...................... 343s ... Sebastian (only 55% GPU load)
GTX 670 (stock) .......................................................... 363s .... zzuupp
GTX 660ti (GPU Boosts to 1058) ................................ 384s .... Sebastian (only 50% GPU load)
GTX 690 (stock) .......................................................... 394s .... S@NL - JBG (dual GPU card)
GTX 750 Ti (stock) ...................................................... 445s ..... TLS
GTX 260 c216 (stock) ................................................. 454s ..... Assimilator1
GTX 650 Ti (stock) ...................................................... 520s ..... fractal
GT 610 (stock) .......................................................... 1827s .... M0CZY

Current CPU statistics ~ Average CPU Time to Complete 1 MW v1.0x 106.88 credit WU :-

Intel i7 5820k (o/c 4.1 GHz) .................................. 2055s ... TennesseeTony
Intel i7 4820k (o/c 3.9 GHz) ................................... 2221s ... Assimilator1
Intel Xeon E5-2695 v3 (o/c 2.8 GHz, 14 cores!) .... 3017s ... TenneseeTony
Intel C2Q Q9550 (o/c 3.6 GHz, 424 MHz FSB) ..... 3159s ... Assimilator1
Intel i7 5820k (o/c 4.1 GHz) ................................... 3221s ... TenneseeTony HT on
Intel Xeon X5570 (Turboing to 3.2 GHz) ............... 3376s ... DutchDK
Intel i7 4820k (o/c 4.34 GHz) ................................. 3597s ... Assimilator1 HT on
Intel i7 4820k (o/c 3.9 GHz) ................................... 3672s ... Assimilator1 HT on
Intel Xeon X5650 (o/c 4.1 GHz) ............................. 4130s ... DR4GOON
AMD Phenom 940 (3 GHz) .................................... 4680s ... TLS
Intel Xeon E5-2695 v3 (o/c 2.8GHz, 14 cores!) ..... 5022s ... TenneseeTony HT on
Intel i7-4702MQ (2.2GHz) ...................................... 6245s ... Maxwell HT on
AMD X2 4600 (2.4 GHz) ........................................ 9780s ... TLS
Pentium D 2.8 GHz ................................................ 9937s ... SharkNose

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

If your wondering why I'm starting over again it's because the original (long) version of the 213.76 credit WU used for the benchmark has become only sporadically available. At 1st a short version joined it & so for a while we had both (which confused some people), then the long 1 disappeared altogether for a time! It is sporadically back but it just makes things too difficult.

Also I later found out that some people don't run the mod fit WUs (like the 213.76s) because of lower credit & some people can't run them due to hardware limitations (e.g HD 3800s). I did ask at the MW forum originally but no one said anything .
So this time I've chosen MW@H v1.02 106.88 credit WU for the benchmark, although it's shorter run times give the illusion of smaller performance gaps between different hardware it'll be easier overall.

Further info.

For reference, link to my 1st MW benchmark thread.
New thread at the Milky Way forums & the KWSN forum.
Threaded added to SETI.USA forum

Stock clocks for some of the above o/ced graphic cards, source Wiki (GPU/RAM MHz):-

HD 5850 ...................... 725/1000
HD 5870 ...................... 850/1200
HD 5970 ...................... 725/1000 (dual GPU)
HD 6950 ...................... 800/1250
HD 6970 ...................... 880/1375
HD 7870 XT ................. 925-975/1500
HD 7950 ...................... 800/1250
HD 7950 Boost ............. 850-925/1250
HD 7970 ...................... 925/1375
HD 7970 GE ................. 1000-1050/1500
R9 280 ........................ 827-933/1250
R9 280X ...................... 850-1000/1500
R9 380X ...................... 970/5700 MT/s
R9 390 ........................ >1000/6000 MT/s
R9 390X ...................... >1050/6000 MT/s

 

[Rudy Toody]

AMD 7900 series: 39.83s average

 

[GLeeM]

106.88 credit WU
V1.02 app
Catalyst 13.12 Win7 64 BOINC 7.2.42

AMD HD7950 (GPU 1,000 RAM 1,250) ................. 28.11 seconds (329,801 ppd)

AMD HD7950 (GPU 1,188 RAM 1,250) ................. 24.12 seconds (384,768 ppd)

 

[Assimilator1]

Thx for the times guys .
Great o/c on that 2nd 7950 Gleem!

Roody
You forgot to tell me exactly which 7900 you have & whether it's at stock clocks or not .
7950 at stock?

 

[Rudy Toody]

Thx for the times guys .
Great o/c on that 2nd 7950 Gleem!

Roody
You forgot to tell me exactly which 7900 you have & whether it's at stock clocks or not .
7950 at stock?

7950 at stock

 

[Assimilator1]

Thx , btw I had noticed your output shoot up recently .
How long you had it?

 

[DutchDK]

Benching mode enabled :
App_config.xml amended to enable only 1 GPU and 1 CPU per WU. HT disabled in bios.
Configuration :
Dual Xeon X5570 CPU, Asus Z8PE-D12X MB, 48 GB DDR3 1333 Mhz ECC ram, 1 Sapphire Vapour-X R9-280X Tri-X OC edition card (factory OC to 1100 Mhz Core/1500 Mhz Mem) running at 1185 Mhz Core/1600 mhz Mem, 1 Asus Radeon DCII-TOP R9-280X card (factory OC to 1070 Mhz Core/1600 Mem) running at 1185 Mhz Core/1600 mhz Mem.
Windows 7 64bit, Boinc 7.2.42(x64), Catalyst 14.4

Times swinging between 23.03 secs and 23.10 secs.
Average time selected in bench of 5 units : 23.04 secs
WUtask WU Computer DateStart DateReported Status Runtime CPUtime Credit Application
754764484 564551494 561358 29 May 2014, 22:01:06 UTC 29 May 2014, 22:33:57 UTC Completed and validated 23.04 1.84 106.88 MilkyWay@Home v1.02 (opencl_amd_ati)
754764482 564551491 561358 29 May 2014, 22:01:06 UTC 29 May 2014, 22:33:57 UTC Completed and validated 23.03 1.78 106.88 MilkyWay@Home v1.02 (opencl_amd_ati)
754763115 564572517 561358 29 May 2014, 21:58:52 UTC 29 May 2014, 22:31:43 UTC Completed and validated 23.04 1.76 106.88 MilkyWay@Home v1.02 (opencl_amd_ati)
754763112 564572514 561358 29 May 2014, 21:58:52 UTC 29 May 2014, 22:31:43 UTC Completed and validated 23.04 1.70 106.88 MilkyWay@Home v1.02 (opencl_amd_ati)
754762465 564572038 561358 29 May 2014, 21:57:44 UTC 29 May 2014, 22:29:30 UTC Completed and validated 23.04 1.90 106.88 MilkyWay@Home v1.02 (opencl_amd_ati)

CPU Xeon X5570 2.93 Ghz Quadcore with Turbo boost up to 3.33 Ghz, normally running Turbo boost at 3.20 Ghz during a Boinc run.
Average per WU : 3,376.028 seconds (56 minutes, 16.028 seconds)

WUtask WU Computer DateStart DateReported Status Runtime CPUtime Credit Application
754730787 564547726 561358 29 May 2014, 21:03:36 UTC 29 May 2014, 22:28:23 UTC Completed and validated 3,381.49 3,364.40 106.88 MilkyWay@Home v1.00
754730176 564547237 561358 29 May 2014, 21:02:29 UTC 29 May 2014, 22:27:16 UTC Completed and validated 3,377.42 3,360.26 106.88 MilkyWay@Home v1.00
754730175 564547236 561358 29 May 2014, 21:02:29 UTC 29 May 2014, 22:27:16 UTC Completed and validated 3,378.44 3,361.87 106.88 MilkyWay@Home v1.00
754730173 564547234 561358 29 May 2014, 21:02:29 UTC 29 May 2014, 22:27:16 UTC Completed and validated 3,365.37 3,348.06 106.88 MilkyWay@Home v1.00
754730148 564547209 561358 29 May 2014, 21:02:29 UTC 29 May 2014, 22:27:16 UTC Completed and validated 3,377.42 3,360.56 106.88 MilkyWay@Home v1.00

 

[zzuupp]

I hope this isn't a double post. The site timed out on my first attempt.

A dedicated CPU core for each GPU.

3

I've got YAFU multicore going. These times don't meet your qualifications.

GTX 670 : 362s. This is a very tight average. All times are 360-365.
GTX 560ti : 264s. This is much looser. I've got high 240s to low 270s with the higher times more prevalent.

I did a scrape this morning at work, and nothing looks much different now.

 

[Rudy Toody]

Thx , btw I had noticed your output shoot up recently .
How long you had it?

Over one year. My production suffered for a while because of mysterious shutdown problem. Ended up being software related to the suspend feature. It looked just like heat failure!

I'm heavy into some mathematics, so I haven't been crunching on any cores, lately.

 

[Assimilator1]

Doh! Nightmare! lol.

Welcome to AT forums DutchDK , awesome rig you've got!
And thx for the times & the detailed info :thumbsup:

zzuup
Sorry I can't use those times then, I want to post times that are possible when the grx card is given optimum resources for an accurate benchmark.
Why don't you just pause (or suspend) the CPU tasks just long enough for your cards to do 5 WUs each? At the most you'd need to do that for 1/2hr, if you deselect the other MW@H apps.

 

[Assimilator1]

Btw does anyone know another Team whose forum & MW team is very active? (so I can gather more times).

Didn't try SETI.USA before because their MW forum looked dead, but looking at the views/replies for the Pentathlon thread maybe I was wrong, so I'm going to try them.

 

[DutchDK]

Welcome to AT forums DutchDK , awesome rig you've got!
And thx for the times & the detailed info :thumbsup:

You'll be even more envious when I tell you what it has cost me.
The CPU's are cooled by a pair of Antec H20-920 v4 AIO watercoolers, the cabinet is a CoolerMaster Cosmos Pure, the PSU is a Corsair 1000 HX.
There are 5 enterprise grade SAS 2 TB drives, 4 running 2 Raid 10 arrays, one dedicated as scratch and temp drive, and a 256 GB Kingston mSATA SSD drive for boot, OS and applications.

All I have payed for, are the two Antec H2O-920 v4 watercoolers, to get rid of the fan noise, and the 5 Corsair Air SP120 High Performance Edition fans there is in the cabinet (2x2 for the Antec radiators, 1 as an exhaust fan at the rear. Existing CoolerMaster A12025-12CB-3BN-F1 120mm fans are reused as bottom intake fan, and HDD enclosure fan) - total noiselevel when idle 32 db(a), total noiselevel when loaded with all cores at 100 % and the two GPU's crunching either Seti@home, PrimeGrid, or Milkyway@home WU's, is nice and comfy 37 db(a), measured at 1 meter from the rig.

Intake and HDD enclosure fans running at 1500-1600 RPM, radiator fans running at 1600 RPM, exhaust fan at 1900-2000 RPM when under load. At idle the revs are respectively 1000, 900, and 1100 RPM.

Temps are nice and comfy as well. CPU1 is at 57C, CPU2 is at 48C (caused by heat from the two graphicscards rising to the area of CPU1), Motherboard and chassis ambient air temp is 38C, harddisks at between 27 and 32C when doing multimedia editing and reencoding.

As I said, all I payed for this was the two Antec units, at 620 DKR a piece (67.50£/113.05$/83.08&#8364. The rest was a gift from my work (it was my old workstation), when they switched workstations to Haswell based rigs.

 

[Rudy Toody]

Btw does anyone know another Team whose forum & MW team is very active? (so I can gather more times).

I don't know if WUprop breaks things down by card. You might check.

 

[Assimilator1]

As far as I can see it doesn't even break down stats between projects , weird 1 that. Thx for the heads up anyway.

DutchDK
Wow that's 1 helluva gift from your work!
They must love you!

 

[DutchDK]

As far as I can see it doesn't even break down stats between projects , weird 1 that. Thx for the heads up anyway.

DutchDK
Wow that's 1 helluva gift from your work!
They must love you!

It does when you select a gpu, or a project - hold your mouse over the text Afficher le menu, and the selection menu pops out.
http://wuprop.boinc-af.org/results/...on+R9+280X&plateforme=win&tri=projet&sort=asc

 

[Assimilator1]

Seem to be a very wide range of times.

No one else want to post any times here?

 

[Assimilator1]

No one else want to post times?

I'm trying to gather times for my C2Q but BOINC refuses to get more CPU WUs for MW , wth??
So far it's only crunched 3 WUs for an average of 3142s.

zzuupp
Been able to run some GPU WUs with a free CPU core?

 

[Assimilator1]

Finally added a time for my 2nd rig's old GTX260 c216, 454s, it's briefly crunching a little MW & POGS before I switch it to F@H for the race (late I know ).

Anyone wanta post some times?

 

[zzuupp]

I can help this time, maybe.
R9 290X with a core free is 43s on milkway. But, it's not sustainable. It locks up after a few hours. (Without a free core, it's widely scattered around 200s.)
I haven't OC'd it; it's running at whatever gigabyte clocked it at.

This is on win7-32. Modified fits aren't supported.

However, on the green box on win7-64. Also, with free cores:

Milkyway on the 560ti is 252, on the 670 it's 363.
For modified fits, the 560ti is 470, and the 670 is 815.

 

[Assimilator1]

Not using the mod fit WUs anymore. Is the 290X factory overclocked? If so maybe it just isn't stable? Have you tried underclocking it to see if it is then ok?

 

[wayliff]

Boinc client
7.2.42 x64
(I tried the newer client 7.4.27 for a few days but my machine was crashing with that)

OS
Windows 7 Pro x64 fully patched

Video Card
Gigabyte Radeon HD 7950 GPU (GV-R795WF3-3GD) Stock (Core 900Mhz)

Driver
AMD 14.100.0.0

Avg Run time
42.809

I used all the WUs (106.88 MilkyWay@Home v1.02) from the first page of valid results.

 

[wayliff]

Boinc client
7.2.42 x64
(I tried the newer client 7.4.27 for a few days but my machine was crashing with that)

Video Card
Gigabyte Radeon HD 7950 GPU (GV-R795WF3-3GD) Stock (Core 900Mhz)

Driver
AMD 14.100.0.0

Avg Run time
42.809

I used all the WUs (106.88 MilkyWay@Home v1.02) from the first page of valid results.

my average is not as good as the other 7950s listed. I wonder why.

 

[StitchExperimen]

Assimilator1 what can you tell me in generalities of how cpu speed effect in percent the outcome?
What have you noticed in other GPU projects?
What have you found from AMD cpus/different models.

 

[Assimilator1]

This is from memory, so IIRC , CPU speed has little impact on GPU crunching times, quite a few people run new-ish GPUs with old-ish CPUs & achieve similar times.

I've not carried out any benchmarks for other projects (bar my own rigs for the F@H bench thread), so can't answer that question.

Re AMD CPUs TBH I didn't get much data on that, their is some in my 1st MW bench thread if you want to check that out.

my average is not as good as the other 7950s listed. I wonder why.

Yours is pretty close to the stock 1 there.

Did you have a dedicated CPU core for it during the benchmark run?
If so then maybe it's the drivers?

zzuupp
Sorted the lock up problem?

 

[wayliff]

Yours is pretty close to the stock 1 there.

Did you have a dedicated CPU core for it during the benchmark run?
If so then maybe it's the drivers?

One CPU is dedicated to that.
I was reading about this GPU and it is factory overclocked, if I'm not mistaken. The core clock is 900Mhz.

I'm going to update the drivers and see if that changes something.