Re: Putting it to the test NVIDIA or AMD for Folding @ Home
2014/04/25 19:07:34
(permalink)
sajinor
That is surprising they are not doing well. The 7970/7950 were brilliant folders for their price (before the mining craze rose prices). I found best performance using beta drivers but running just "client-type" "advanced" or no flag. It doesn't seem to download the beta core. I have only had experience with 7970/7950 so far.
I should have some R9 290's (non X) coming, unless the seller backs out (Joys of eBay). I will be able to have a closer example at that point.
Core 16 was actually worse than core 15, in points difference from core 17, but if i recall correctly, it paved the way for all the openCL optimizations we see now.
On "openCL optimizations" do I need to install the Intel SDK for openCL for any AMD Card? Thanks
|
Re: Putting it to the test NVIDIA or AMD for Folding @ Home
2014/04/25 19:14:33
(permalink)
Drazhar
bcavnaugh
I just bought 2 PowerColor LCS AXR9 290X 4GBD5-WMDH/OC Radeon R9 290X 4GB 512-Bit GDDR5 PCI Express 3.0 HDCP to put to the test AMD for Folding or NVIDIA. I will add them into the X79 Classified Motherboard with the i7-3930K CPU and 16 GB of memory. CPU Motherboard and GPU all under Water.
My compare will be between the 2 GTX 780 (i7-3970X) and then 2 GTX Titans (i7-4960X) All on Water and on X79 Motherboards.
Testing starts next Weekend.
Huh, 290x vs regular 780 and Titan instead of 780 Ti and Titan Black or whatever the Titan fullGK110 is?
I'm interested, but I thought the competition was 290x and Ti.
So then the 290X is to the Ti and then the 290 is to the 780. It is more of what GPU will Fold What Cores and Projects best. Not looking at the gamming side at all. So for all around I see Newer NVIDIA Card out Folds a Newer AMD Card. But I still have a lot of testing to complete before I make up my mind on which card is better for Folding & Crunching And it gives me something to do on the weekends to take my mind off of work.
|
sajinor
ACX Member
- Total Posts : 375
- Reward points : 0
- Joined: 2008/03/29 22:00:57
- Location: Oregon
- Status: offline
- Ribbons : 2
Re: Putting it to the test NVIDIA or AMD for Folding @ Home
2014/04/25 19:42:42
(permalink)
bcavnaugh
On "openCL optimizations" do I need to install the Intel SDK for openCL for any AMD Card?
Thanks
I believe everything you need is in the AMD driver. At least it was in my case. The box running the stats for the TZC has a 7970 in it, and has nothing installed but MSI Afterburner, the AMD driver, and Google Chrome. Oh yeah, if you do want to tinker with overclocking on AMD, MSI Afterburner will probably be the way to go. It is free, works with most cards, and functions basically the same as Precision.
 aF: 4x 8425 HE @ 2.1, H8QMEIBM dx360 B1: 2x L5640 @ 2.53; B2: 2x L5640 @ 2.53Kamehameha: i7-4790K @ 4.4, Antec Kuhler 620, EVGA Z87 Classy, 8GB DDR3 2133, 3x R9 290x, Antec Eleven Hundred V2Primary: i5-3570k @ 4.6, Swiftech H240x Opened, EVGA Z77 FTW, 8GB DDR3 2400, GTX 780 Classified, Fractal Define XL 2
|
Re: Putting it to the test NVIDIA or AMD for Folding @ Home
2014/04/26 08:16:04
(permalink)
Both Rigs are now set with client-type = advanced and the both pulled P9101 MUCH Better then P9102 Beta. As you can see the AMD is now Folding Better Than the GTX 780....More Projects Will Only Tell. Status Progress Name TPF PPD WUs ETA Core ID Project (Run, Clone, Gen) Credit
Running 18% LCS R9 290X Slot 00 00:02:13 180385 6.496 03:01:46 ZETA 17 P9101 (R901, C0, G38) 27767.6
Running 18% LCS R9 290X Slot 01 00:02:15 176638.7 6.4 03:04:30 ZETA 17 P9101 (R975, C0, G30) 27599.8
Running 15% GTX 780 Slot 00 00:02:37 139674.6 5.503 03:42:25 ZETA 17 P9101 (R508, C0, G55) 25380.7
Running 8% GTX 780 Slot 01 00:02:44 130939.6 5.268 04:11:28 ZETA 17 P9101 (R37, C0, G48) 24854.3
Running 21% GTX 780 Slot 02 00:02:39 136951.1 5.434 03:29:21 ZETA 17 P9101 (R96, C0, G30) 25202.8 Two GTX 780 PPD 276625.7 Credit 55367.4 to the Two R9 290X PPD 357023.7 Credit 50583.5 Progress Name TPF PPD WUs ETA Core ID Project (Run, Clone, Gen) Credit
18% LCS R9 290X Slot 00 00:02:13 180385 6.496 03:01:46 ZETA 17 P9101 (R901, C0, G38) 27767.6
18% LCS R9 290X Slot 01 00:02:15 176638.7 6.4 03:04:30 ZETA 17 P9101 (R975, C0, G30) 27599.8
Totals 357023.7 12.896 Totals 55367.4
15% GTX 780 Slot 00 00:02:37 139674.6 5.503 03:42:25 ZETA 17 P9101 (R508, C0, G55) 25380.7
21% GTX 780 Slot 02 00:02:39 136951.1 5.434 03:29:21 ZETA 17 P9101 (R96, C0, G30) 25202.8
Totals 276625.7 10.937 Totals 50583.5
AMD WINS OUT Deff: 80398 Deff: 4783.9
post edited by bcavnaugh - 2014/04/26 08:26:32
|
Orange_1050
FTW Member
- Total Posts : 1181
- Reward points : 0
- Joined: 2011/10/10 00:48:32
- Status: offline
- Ribbons : 3
Re: Putting it to the test NVIDIA or AMD for Folding @ Home
2014/04/27 12:30:01
(permalink)
bcavnaugh, I'm sorry not to have commented this fantastic Project sooner. I've been away for almost a week, only had some TeamView sessions to maintenance some of my Troopers. See, in Norway, it's quite common to have a Summer house, and so do I. I Guess you know all about maintanance on buildings and machines, well I wear hard labor clothing all my spare time  Taking Down trees, painting, carpenting, gardening and so on. Wonder how I have time to crunch? ha ha. Anyways, Your rig is a true beauty my friend..... Nuff said... 
  My Virgin Prime: 433*2^2188076+1 (658680 digits)
|
Re: Putting it to the test NVIDIA or AMD for Folding @ Home
2014/04/27 15:31:38
(permalink)
Thanks Orange_1050 I am still testing other Projects. I did find that with Collatz Conjecture that you can set a GPU to CPU of 1 to 1 or what ever you want. <app_config> <app> <name>large_collatz</name> <max_concurrent>2</max_concurrent> <gpu_versions> <gpu_usage>1.0</gpu_usage> <cpu_usage>1.0</cpu_usage> </gpu_versions> </app> </app_config> max_concurrent is the number of GPU Cards you have installed. cpu_usage as shown above it 1 CPU for 1 GPU Like on my GTX 690 I set the cpu_usage to 0.5 This is my Post: application Large Collatz Conjecture needs 56 Days to complete Still working settings in the "large_collatz_6.04_windows_x86_64__opencl_amd_gpu.config" file. verbose=1 items_per_kernel=22 kernels_per_reduction=9 threads=8 sleep=1 PS NEVER SET <cpu_usage>1.0</cpu_usage> to 2.0 BSOD Heaven
post edited by bcavnaugh - 2014/04/27 16:05:23
|
Re: Putting it to the test NVIDIA or AMD for Folding @ Home XOR Crunching Now
2014/05/09 09:58:35
(permalink)
I am running Collatz Conjecture (solo_collatz) on my NVIDIA GPU's and time for each is between 54 minutes (Titans (CPU 4.3)) and 1 Hour 5 minutes for the GTX 680 (CPU 4.3) Classifieds and GTX 690 (CPU 4.2). NVIDIA <app_config>
<app>
<name>solo_collatz</name>
<max_concurrent>3</max_concurrent>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config> On the LCS R9 290X Over 2 Hours (CPU 4.6) I do not plan to keep it this high but for testing I really wanted the AMD to do better. AMD <app_config>
<app>
<name>solo_collatz</name>
<max_concurrent>2</max_concurrent>
<gpu_versions>
<gpu_usage>1.0</gpu_usage>
<cpu_usage>1.0</cpu_usage>
</gpu_versions>
</app>
</app_config> All GPU's are at Stock Settings None are Overclocked or Over Voltage and No K-BOOST Enabled.
post edited by bcavnaugh - 2014/05/09 10:14:20
|
Re: Putting it to the test NVIDIA or AMD for Folding @ Home XOR Crunching Now
2014/05/09 11:23:36
(permalink)
Solo Collatz Time: 1 Hour Credit: 14,549.33 GPU: GTX 690 6556225 6021151 9 May 2014, 15:07:43 UTC 9 May 2014, 16:47:20 UTC Completed and validated 3,952.56 1.73 14,549.33 Solo Collatz Conjecture v6.04 (cuda55) Large Collatz Time: 17 Hours Credit: 236,943.44 GPU: GTX 690 6160720 5724777 30 Apr 2014, 20:34:14 UTC 9 May 2014, 18:17:15 UTC Completed and validated 62,436.64 2.32 236,943.44 Large Collatz Conjecture v6.04 (cuda55) Doing what I can for now. Collatz Conjecture Crunching@EVGA
(Down 1 places1) 81
9,461.279461 (Down 6.734007 points6.734007)
More data to Follow....
post edited by bcavnaugh - 2014/05/09 14:05:09
|
Re: Putting it to the test NVIDIA or AMD for Folding @ Home XOR Crunching Now
2014/05/23 20:43:49
(permalink)
For the most part the AMD get about 10,000 Credits per GPU compared to the GTX 780 on at least Core 17 P13000/P13001 Projects. Or 207403.9 (2.462 WUs) for the AMD and 150545.4 (1.991 WUs) for NVIDIA PPD TPF 00:05:52 AMD & TPF 00:07:26 NVIDIA Running Current Drivers on both AMD and NVIDIA AMD Cross Fire Off as well as NVIDIA SLI is Off For ATM that is AMD is the better GPU for Folding, Sad I know and I have so much invested in NVIDIA for Folding. So for future Folding and Crunching I will keep the two AMD GPU's Folding and move ALL my other Rigs over to Crunching, for the most part I have done this. The Odd part with all my testing over the last month shows that with Crunching both cards do about the same. I see no point in posting my finding on https://foldingforum.org I have had no real help from them and they really do not support single folders at lest that is what I have found. Maybe next Fall it will get better over their and we might have a good Folding Driver from NVIDIA. Keep On Folding & Crunching Teams EVGA.
post edited by bcavnaugh - 2014/05/23 21:01:20
|
Re: Putting it to the test NVIDIA or AMD for Folding @ Home XOR Crunching Now
2014/05/23 21:07:22
(permalink)
End Of Testing: Status Progress Name TPF PPD Wus ETA Core Core ID Project (Run, Clone, Gen) Credit
Running 2% GTX 780 On Slot 00 00:07:27 144403.0 1.933 12:10:06 ZETA 17 P13000 (R173, C0, G13) 74708.5
Running 3% GTX 780 On Slot 01 00:07:30 142927.5 1.920 12:07:30 ZETA 17 P13001 (R94, C7, G7) 74441.4
Running 16% GTX 780 On Slot 02 00:07:14 150841.6 1.991 10:07:36 ZETA 17 P13001 (R16, C9, G6) 75770.0
Running 69% LCS R9 290X On Slot 00 00:05:52 206145.2 2.455 03:01:52 ZETA 17 P13001 (R39, C5, G9) 83985.1
Running 69% LCS R9 290X On Slot 01 00:05:52 206145.2 2.455 03:01:52 ZETA 17 P13001 (R279, C4, G13) 83985.1 Name TPF PPD Wus Credit
GTX 780 00:07:24 146057.4 1.948 74973.3
LCS R9 290X 00:05:52 206145.2 2.455 83985.1 Chart to Follow: 
post edited by bcavnaugh - 2014/05/23 21:27:59
Attached Image(s)
|
bill1024
CLASSIFIED Member
- Total Posts : 3252
- Reward points : 0
- Joined: 2008/10/18 01:01:10
- Status: offline
- Ribbons : 12
Re: Putting it to the test NVIDIA or AMD for Folding @ Home XOR Crunching Now
2014/05/23 22:03:28
(permalink)
Hey Bill, any idea what cpu has the best cost of electric, points per watt?
I have always likes AMD cards. I used to think the color was more vibrant.
Biggest problem I remember was driver issues. Run this driver got this game and that driver for that game.
Life is too short to carry a cheap pocket knife. CaseXX, Al Mar, Hubertus, Frank b, A\B, SchattMorgan U25ITA93JV 
|
BadBertie
SSC Member
- Total Posts : 599
- Reward points : 0
- Joined: 2012/08/10 17:09:48
- Location: France
- Status: offline
- Ribbons : 0
Re: Putting it to the test NVIDIA or AMD for Folding @ Home XOR Crunching Now
2014/05/24 00:21:28
(permalink)
Great research there Bill. So thanks for sharing.
You aren't the only one disappointed Nvidia came in 2nd but it's important to challenge accepted standards from time to time.
Stop us getting stuck in a rut.
Maybe its a one-off with this current range of AMD cards or an indication of improved driver performance by AMD for folding.
In any case (pardon the pun) it's all good knowledge.
Maybe Nvidia will step it up again as a result.
#1- Switch 810 white. P8z68 Deluxe gen 3. i5 2500k 4.8ghz. 16gb Corsair Vengeance. 3 x EVGA gtx 670. Silencer mkIII 1200W. EK dual cylinder res. Alphacool Nexxxos ut60 3.120 & 2.120 rads. #2- H8QGi+F. 4P 6274. 64gb ddr3. Corsair HX1050. #3- H8QGL-iF 4P 6172. 32gb ddr3 #4- Inside-out project ... Tyan S8238 2P 6172 ...  
|
linuxrouter
Omnipotent Enthusiast
- Total Posts : 8043
- Reward points : 0
- Joined: 2008/02/28 14:47:45
- Status: offline
- Ribbons : 102
Re: Putting it to the test NVIDIA or AMD for Folding @ Home XOR Crunching Now
2014/05/24 19:25:59
(permalink)
Thanks for sharing your results. I recently purchased a R9-290x. The R9-290x with a core frequency of 1075 MHz is folding well at approximately 231K PPD with p9102. My EVGA 780 Ti with a boost frequency of 1137 MHz has a PPD ranging from 219-245K with p9406. The load temperature of the R9-290x is approximately 74C compared to 73C of the 780 Ti. Both cards are running using a more aggressive fan profile set with Afterbuerner and Precision respectively. Unfortunately, the R9-290x does not seem to perform as well with other OpenCL distributed computing applications. I have been testing for days now in Windows and Linux OS and my 7970 cards are out running the R9-290x card with applications including Einstein, Milkyway, and Seti. With Milkyway running a FP64 based application, the performance difference makes sense. However, with the other two projects that primarily run FP32 applications, the R9-290x should be able to outrun the 7970 which has not been the case.
post edited by linuxrouter - 2014/05/24 20:36:34
CaseLabs M-S8 - ASRock X99 Pro - Intel 5960x - XSPC WC - EVGA 980 Ti Hybrid - Samsung 950 512GB - EVGA 1600w TitaniumAffiliate Code: OZJ-0TQ-41NJ
|
Re: Putting it to the test NVIDIA or AMD for Folding @ Home XOR Crunching Now
2014/05/24 23:05:18
(permalink)
bill1024
Hey Bill, any idea what cpu has the best cost of electric, points per watt?
I have always likes AMD cards. I used to think the color was more vibrant.
Biggest problem I remember was driver issues. Run this driver got this game and that driver for that game.
Well If looked at my 4P and 2P AMD both I would have to say Intel By Far time to complete projects, but for cost per watt AMD. One item I have not learned is to OC the AMD CPU's I have no real issues OC Intel CPU's at least the i7 for Socket 2011.
|
Re: Putting it to the test NVIDIA or AMD for Folding @ Home XOR Crunching Now
2014/05/24 23:08:26
(permalink)
linuxrouter
Thanks for sharing your results.
I recently purchased a R9-290x. The R9-290x with a core frequency of 1075 MHz is folding well at approximately 231K PPD with p9102. My EVGA 780 Ti with a boost frequency of 1137 MHz has a PPD ranging from 219-245K with p9406. The load temperature of the R9-290x is approximately 74C compared to 73C of the 780 Ti. Both cards are running using a more aggressive fan profile set with Afterbuerner and Precision respectively.
Unfortunately, the R9-290x does not seem to perform as well with other OpenCL distributed computing applications. I have been testing for days now in Windows and Linux OS and my 7970 cards are out running the R9-290x card with applications including Einstein, Milkyway, and Seti. With Milkyway running a FP64 based application, the performance difference makes sense. However, with the other two projects that primarily run FP32 applications, the R9-290x should be able to outrun the 7970 which has not been the case.
You know I have not set mine to frequency of 1060 as that is what they are set to run at but they are set to only 1000. All my cards are on Water and so far my max temp was 50C. Thanks
|
