Hello, to add to the new list :
Mean over the last 20 validated units : 127.02s
Config : Phenom II 1100T stock clocks (3.3GHz) + RX 470 4GB (1205MHz GPU, 1850MHz RAM)

Hello, to add to the new list:
Mean over the last 20 validated units: 169.75 secs
Config: AMD Ryzen 1700X @ 3.9 Ghz + GTX 1070Ti @ 2.0 Ghz GPU, 8.4 Ghz RAM

---

DVDL
Is that time from at least 5 WUs?

Yes, i've done a ton of them. Not anymore, hope to get an AMD card sometime.

Sorry, new question ;), were they all from 227.2x results?

Yes, correct.

you didn't add my 6970 to your list!

it takes 94 second per WU

AMD Phenom(tm) II X6 1090T (6 cores)

Thanks Keith & AF HFR Seeds, I will add your times :). Added to AnandTech table.

DVDL
Sorry, only just noticed you didn't say what CPU that was running with?

you didn't add my 6970 to your list!

it takes 94 second per WU

AMD Phenom(tm) II X6 1090T (6 cores)

As per my answer on the 26/9/17 http://milkyway.cs.rpi.edu/milkyway/forum_thread.php?id=3551&postid=66666, I was awaiting an answer from you! As to whether the time was from an average of at least 5 WUs.
Which you still haven't answered ;)

And I assume your CPU isn't o/ced?

---

Team AnandTech - SETI@H, DPAD, F@H, MW@H, A@H, LHC, POGS, R@H, Einstein@H, DHEP, WCG

Main rig - Ryzen 5 3600, MSI B450 G.Pro C. AC, RTX 3060Ti 8GB, 32GB DDR4 3200, Win 10 64bit
2nd rig - i7 4930k @4.1 GHz, HD 7870 XT 3GB(DS), 16GB DDR3 1866, Win7

yes it is the average of more than 5 WU.

Both my cpu and gpu are not OC.

Ok cool, thanks, added your time to the AT table :).
Will post a new table here when DVDL answers.

---

Team AnandTech - SETI@H, DPAD, F@H, MW@H, A@H, LHC, POGS, R@H, Einstein@H, DHEP, WCG

Main rig - Ryzen 5 3600, MSI B450 G.Pro C. AC, RTX 3060Ti 8GB, 32GB DDR4 3200, Win 10 64bit
2nd rig - i7 4930k @4.1 GHz, HD 7870 XT 3GB(DS), 16GB DDR3 1866, Win7

Hi guys,

Here is my benchmarks for a GPU I pulled out of my HTPC, an AMD HD 7750 and an R9 280X.

646.33
646.79
646.94
647.57
647.03
Ave: 646.93

For my R9 280X
51.68
50.94
48.20
50.06
51.02
Ave: 50.38

Both GPUs benefit from doing 2 tasks at a time: with the HD7750 getting 1252.74 (626.37 per task, not a lot) and the R9 280X at 70.62 (35.3 per task).

I will post my CPU results after they are completed: Ryzen R7 1700X stock with 2667 MHz memory.

Ok, I see I left some information out. The AMD 280X is stock (1030 MHz clock, 1500 MHz memory). The HD 7750 is stock.

Windows 10
AMD driver: Crimson Relive 17.11.1
Boinc: 7.8.3 (x64)

CPU: AMD R7 1700X stock 3.4
Memory overclocked to 2667 MHz
16 GB memory

No hyperthread
3213.86
3282.54
3383.62
3354.03
3327.06
Ave: 3,315.22 (55.25 min/unit) 8 cores

Hyperthread
4423.29
4428.60
4433.04
4425.45
4427.59
Ave: 4,427.67 (73.79 min/unit) 16 cores

A couple more computers:
Windows 10
AMD driver: Catalyst 15.7.1
Same Boinc version.
8 GB 1600 MHz memory

AMD 8350
5156.96
5121.07
5211.27
4779.63
4806.48
Ave: 5105.08 (83.58 min/unit) 7 cores

AMD HD 5870 (900 MHz core, 1050 MHz memory)
116.06
116.06
116.13
116.06
116.06
Ave: 116.07 (1.93 min/unit)

Another AMD 8350
5452.37
5467.93
5485.56
5278.12
5257.72
Ave: 5388.34 (89.8 min/unit) 7 cores

AMD HD 6970 (880 MHz core, 1375 MHz memory)
107.33
107.35
107.44
107.62
106.86
Ave: 107.32 (1.79 min/unit)

Hope I got all the info you need! This has been an interesting thread.

Joe

LOL units of measure in seconds for the 280X and HD 7750. I know you know that but I should have put that in.

DVDL
Sorry, only just noticed you didn't say what CPU that was running with?

Pentium G3900

Hmm hope I get this right...(a lot of variables)

CPU -2500K (O/C to 4.3Ghz)
RAM - 8gig DDR3
Windows 7 64 Bit
BOINC version - 7.8.3
Video Card - Vega 56 stock.
GPU Driver - 17.11.1

Average over 5 WU's - 63sec. (315sec)

Now where I get a bit confused is I'm using this app_config for the test.
I still set CPU usage to 75% though.

<app_config>
<app>
<name>milkyway</name>
<max_concurrent>12</max_concurrent>
<gpu_versions>
<gpu_usage>1</gpu_usage>
<cpu_usage>0.2</cpu_usage>
</gpu_versions>
</app>
<app>
<name>milkyway_separation__modified_fit</name>
<max_concurrent>12</max_concurrent>
<gpu_versions>
<gpu_usage>1</gpu_usage>
<cpu_usage>0.2</cpu_usage>
</gpu_versions>
</app>
</app_config>

Seems a bit slow really. I'm currently running 4WU's with "Use at most 25% of CPU", or I run 3WU's at 50%. Works out about 49 - 50sec for the WU's.

Let me know if I should change something and I'll run it again.

---

Thanks Chooka :)
I've never heard of a Vega 56, lol. I see they only came out in August, I've got some reading to do! Looking at the wiki for it https://en.wikipedia.org/wiki/AMD_RX_Vega_series , I see the 56's DP when boosting is 659, & 518 GFLOPS otherwise. Which is somewhere between the HD 6930 & the HD 6970, above & below the R9 290, & when boosting just about the R9 390 (for DP only, which is what MW@H uses).

I see it's not a cheap card btw! :Q

JoeM
Thanks for all the times, lots of info :), but you still missed off what CPUs those GPUs run with ;) lol.
Interesting to see even the HD 7750 benefits from running 2 WUs at once!
Btw 1030 MHz for the 280X isn't AMD stock, I guess it must be a factory o/c. You've just pipped me to the top ;).
Are the 8350s @ stock?
What CPU does the HD 6970 & 5870 run with?
Any differences between the 2 8350s to account for the slightly different times? Clock speed?

DVDL
Cheers :)

wb8ili
I appreciate the times you've given me but there's a bunch of information missing, were the times all derived from an average of a min of 5 WUs?
What CPU's were they running with?
Were either the CPUs or GPUs overclocked? if so, what to?

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

Folks, I'm spending a load of time on follow up questions to requirements asked for in the table. And I don't have so much spare time here anymore ;).

Please could anyone who wants there time posted to the table, state at least :-
GPU model & speed, or state 'stock'.
State what CPU the GPU was running with & whether it's o/ced (CPU type & speed has been affecting GPU WU times now for over a year).
Confirm that the time is an average of at least 5 227.23 credit WUs.

For the best & most stable times you need a free physical CPU core for your GPU, & for the GPU to be crunching only 1 WU at a time.

Btw, I've dropped the requirement to state GPU RAM MHz as it makes no real difference, but feel free to state it as people can hit the link to your post to look for more info :).

Adding times to the AT table now.....

---

Team AnandTech - SETI@H, DPAD, F@H, MW@H, A@H, LHC, POGS, R@H, Einstein@H, DHEP, WCG

Main rig - Ryzen 5 3600, MSI B450 G.Pro C. AC, RTX 3060Ti 8GB, 32GB DDR4 3200, Win 10 64bit
2nd rig - i7 4930k @4.1 GHz, HD 7870 XT 3GB(DS), 16GB DDR3 1866, Win7

Chooka
You got it right 1st time :).
Nice time for a modern card btw :)

Darrell
What speed was that 1100T running at? Stock?

Latest benchmark table

Since v1.46 was released on 1/5/17 (UK date format :p), the WU times & credits changed. Times are apparently 'slightly longer' & the main WUs (99%+) & thus the new benchmark WU is 227.23 credits. See below the v1.46 table for the other benchmark requirements.

Btw, watch out for the 227.26 credit WUs, they are very rare (approx. 1% of WUs atm), but despite their tiny increase in credit they take about 5% longer, at least on my HD 7970, ~56s vs 53s.

Current GPU statistics ~ Average Run Time to Complete 1 MW v1.46 227.23 credit WU :-

R9 280X, GPU 1030 MHz (CPU, ???) .................................................................. 50.4s .... JoeM
HD 7970, GPU 1000 MHz (CPU, i7 4930k @4.1 GHz) ......................................... 53s ....... Assimilator1
Vega 56, stock (CPU, 2500k @4.3 GHz) ............................................................... 63s ....... Chooka
HD 6970, GPU 890 MHz (CPU, Phenom II X6 1090T, stock) .............................. 94s ....... Hassan Shebli
HD 6970, stock (CPU, ???????) .......................................................................... 107s ....... JoeM
RX 480 8GB, GPU o/c to? (CPU, Phenom II X6 1100T @?) ............................... 110s ....... Darrell
HD 5870, GPU 900 MHz, (CPU, ?????? ) ........................................................... 116s ....... JoeM
RX 470 4GB, GPU 1205 MHz (CPU, Phenom II 1100T, stock) ......................... 127s ....... [AF>HFR] Seeds
GTX 1070 Ti, GPU 2 GHz (CPU, Ryzen 1700X @3.9 GHz) ................................ 169.8s .... Keith Myers
GTX 1060, stock (CPU, Pentium G3900) ............................................................ 250s ...... DVDL
HD 7750, stock (CPU, ? ) ..................................................................................... 647s ........ JoeM


Current CPU statistics ~ Average Run Time to Complete 1 MW v1.46 227.23 credit WU :-

Ryzen R7 1700X (8C, stock 3.4 GHz, RAM o/c 2667 MHz) .................................. 3315s no HT ... JoeM
Ryzen R7 1700X (8C, stock 3.4 GHz, RAM o/c 2667 MHz) .................................. 4428s HT on ... JoeM
8350 (7C, ?????) ................................................................................................... 5105s ...... JoeM
8350 (7C, ?????) ................................................................................................... 5388s ...... JoeM

For links to each time go to the AnandTech thread here https://forums.anandtech.com/threads/milkyway-h-benchmark-thread-winter-2016-on-different-wu-gpu-cpu-times-wanted.2495905/#post-38654842

---

Team AnandTech - SETI@H, DPAD, F@H, MW@H, A@H, LHC, POGS, R@H, Einstein@H, DHEP, WCG

Main rig - Ryzen 5 3600, MSI B450 G.Pro C. AC, RTX 3060Ti 8GB, 32GB DDR4 3200, Win 10 64bit
2nd rig - i7 4930k @4.1 GHz, HD 7870 XT 3GB(DS), 16GB DDR3 1866, Win7

Nvidia 1080Ti avg of last 5 wu's 124.154

CPU type AuthenticAMD
AMD FX(tm)-8350 Eight-Core Processor [Family 21 Model 2 Stepping 0]
Number of processors 8
Coprocessors NVIDIA GeForce GTX 1080 Ti (4095MB) driver: 385.69 OpenCL: 1.2
Operating System Microsoft Windows 7
Ultimate x64 Edition, Service Pack 1, (06.01.7601.00)
BOINC version 7.8.3
Memory 12188.16 MB
Cache 2048 KB

No overclocking, yes I have one cpu core reserved for use by the gpu.

Hi Assimilator,

Sorry I wasn't so clear in that post about the two 8350s (message 66799). Both the HD5870 and the HD6970 were mated with the 8350 CPUs. The HD5870 mated with the slightly faster one. Both 8350 CPUs are running stock, no overclock.

I suspect that the motherboards and memory might make a difference in the two computers (CPUs) but I haven't investigated (both are different brands). It is unlikely that the CPUs were throttling since both are liquid cooled and runs in the mid 50s to around 60 max.
If I get some time over the holidays, I may look into the differences.

Hope you have a great holiday season and hope everyone crunching has one too!!

Let me know if missed something again ;)

Oh I see I didn't post the CPU for the R9 280X as well. It is the Ryzen 1700X stock with 2667MHz memory.

CPU for the HD7750 was also the Ryzen 1700X.

Here are some results from a 7970 running at stock clocks in Linux.

OS: Manjaro Linux kernel 4.15
BOINC Version: 7.8.4
GPU Driver: Catalyst 15.9
CPU: AMD R7 1700 @ 3800MHz
GPU: AMD HD7970 @ 925/1325

10 WU avg run-time - 53.9s

CPU 1700x 3200mhz ram @ 3850mhz: GPU 1 290xt @ 1050mhz 4gb @ 1350mhz ram: 110s per two WU, .25 CPU use, or 65s single WU.

Same system:

GPU 1 7970 @ 1050mhz 3gb @ 1425mhz: 77s per two WU, .25 CPU use, or 45s single WU.


AMD users use two WU!!!

---