• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Anyone planning to test DC projects on AMD Ryzen?

pututu

pututu

[H]ard DC'er of the Year 2021
2FA
Joined
Dec 27, 2015
Messages
3,533
Over the last few days, so much news about Ryzen 1800X beating Intel i7-6900K in non DC testing, mostly gaming. Power dissipation and cost look great. If anyone is planning to get one, would appreciate sharing the test result. Would be interesting to see how it performs in PrimeGrid-LLR, WCG, etc

If the consumer chip outperforms intel in power-performance arena, I'm guessing good showing will be on their server chips too. Will see.
 
I am waiting to see what [H] says on Ryzen. If they do indeed out-perform Intel's flagship CPUs for half the cost then I will be getting at least one build to test it with WCG. Although I am leaving for vacation on the 1st and wont get back until the 14th. So it'll be mid-end of March before I have one ready to go.

Again this depends on real-world reviews from [H] and not the marketing hype BS that AMD is presenting us. This could very well be Bulldozer 2.0.
 
I'm waiting for the 2p naples based workstations to come out before I decide what to do, some of my rigs need upgrades but core A7 is making cpu folding viable again. When the wife gets back to work i'm going to see just what you can expect from core A7 and start making some decisions.
 
If Ryzen lives up to its claim, I think Naples will be awesome in price and performance and likely to be cheaper than xeon v4/v5.... Anxiously waiting to see some numbers posted here or elsewhere soon.

I also read that GPU 1080Ti release is likely to be confirmed next week and if it is, I'm sure one of the crunches/folders will rush to get one and post some benchmark results.

Exciting times...
 
I intend to replace my SLI 970 main gaming rig with a pair of 1080 Ti rigs given the performance boost above 980 Ti/1080 is worth it. Really want to go ahead and get a 4K gaming monitor to start playing on also.
 
I really hope AMD has improved their implementation of the AVX instruction set. Some how they missed the mark in the past.
Would love to see what these CPUs do with primegrid LLR tasks. If they get it right I may build an AMD system. Its been quite a while.
Either way it is good for the consumer Intel gets some competition, Nvidia too for that ,matter. Hopefully lower priced with more of an increase in performance.
 
Bill1024 said:
I really hope AMD has improved their implementation of the AVX instruction set. Some how they missed the mark in the past.
Would love to see what these CPUs do with primegrid LLR tasks. If they get it right I may build an AMD system. Its been quite a while.
Either way it is good for the consumer Intel gets some competition, Nvidia too for that ,matter. Hopefully lower priced with more of an increase in performance.
Click to expand...
Can't confirm this but according to this website, Ryzen needs 2 cycles to run AVX instruction (128-bit) versus 1 cycle for intel with AVX (256-bit). AVX2 has 512-bit instructions. In anyway, I'm hoping someone here or elsewhere with Ryzen could share their findings soon but nevertheless Ryzen supports AVX and AVX2 (at least from the software side) from what I have read.
 
I pre ordered a 1700, I'll post numbers as soon as it gets here for WCG/F@H. Wonder what the AVX issue will cause in terms of F@H and WCG, do they use those instructions?
 
lelliott731 said:
I pre ordered a 1700, I'll post numbers as soon as it gets here for WCG/F@H. Wonder what the AVX issue will cause in terms of F@H and WCG, do they use those instructions?
Click to expand...
I know PrimeGrid LLR projects use AVX instructions.

1700 has 8 cores with base clock of 3GHz. I have xeon E5-2670 with 8 cores running at 3GHz turbo. My rig has ECC (only three DDR3 channels working) with Windows 7, maybe we can compare the performance when you have your build ready. Though not a best comparison but if others would like to chime in and help, that would be great.
 
Ok cool, we will compare numbers. So then the new A7 core will operate at half the speed of other cores?
 
WCG will not take advantage of AVX as they don't have optimized applications. WCG doesn't do any of the coding/compiling and so unless the scientists put forth the effort... you get a universal application and hope for the best. This is why AMD chips are typically brought over to WCG. It is one of the few projects still "worth" running them on.
 
Oh ok, didn't know that. Hopefully the Ryzen will really scream then for WCG!
 
lelliott731 said:
Ok cool, we will compare numbers. So then the new A7 core will operate at half the speed of other cores?
Click to expand...

It shouldn't do, A7 PPD on an AVX enabled CPU can be excellent, people have reported 100-150k from 6950k and i've seen 230k+ from a pair of ivy bridge based xeons so qwhilst not bigadv territory, certainly better than the a4 points we are used to seeing.
 
Yeah, kind of bummed about the gaming. But considering I do mainly multi threaded work, I still like it! Plus the cost is great compared to a equivalent Intel 8 core.
 
Saw a post on Ryzen 1800X doing FAH A7 with about 105K PPD. Personally, GPU would be a better choice for this project if you do have a decent card like 950/1050, I think.

Still searching for WCG numbers other than this but no luck yet even in the WCG forum. Only hardware discussion. Probably nothing to bragg about....:notworthy:
 
I'm still waiting on my motherboard, have a whole system waiting, really getting irritated.
 
Got my Ryzen 1700X system together and running yesterday. Because of the lack of available motherboards I had to settle for a B350 board. Haven't moved my GTX 1080 into it yet. Making sure it's stable before doing so. The BIOS seems to be a WIP. Had to update BIOS just to get Win 10 installed from a USB stick. Installed BOINC and started running WCG on it.
 
Go to the WCG web site. Click on My Contribution then Device Stats. You'll see a list of all your devices. Click on the AMD one and it'll break down the points by day.
 
This?

Statistics Date
white.gif
Total Run Time (y:d:h:m:s)
white.gif
Points Generated
white.gif
Results Returned
03/21/2017
white.gif
0:008:12:49:51
white.gif
48,444
white.gif
143

Just the first full day.

Probably should add that I don't pick and choose projects based on the PPD they give. I do them all.
 
Last edited:
Ok; too bad. The term PPD jumped into my eyes ... thought it was FAH-exclusive.

Enjoy your Ryzen
 
Correct. It's from the devices page at WCG. BOINC doesn't provide any way I know of to seperate the credit for my different machines.
 
Just checked the WCG site again and it now shows: 03/21/2017
white.gif
0:010:14:07:19
white.gif
59,522
white.gif
167
I only upload and download wus between midnight and 5am. Not sure why stats are changing during the day. WUs getting validated maybe.
 
Is the rig running WCG exclusively on all cores?

~50k points is pretty damn good. My 10 Core (20 thread) Intel rig gets around 90k points per day, but it's crunching on 18 of 20 threads (2 threads are dedicated to the two GPUs for F@H)
 
I would suggest we select one WCG project that is pretty consistent in terms of points per hour awarded. SCC appears to be one and generally has quorum of one if you crunch longer period (like a day or more). Then I would pull 20-30 data points or more and calculate the average points per hour per logical CPU. It appears to be pretty consistent from WU to WU. Multiply this number by number of cores and by 24 hrs to give you average PPD.

Step 1: Click on "My Contribution" which is located on the top, then click on "Result Status", underneath "My Projects"
upload_2017-3-21_21-46-53.png


Step 2: Choose the "Device Name" and "Project Name" i.e. SCC. If needed select "valid" from the "Result Status" field.

upload_2017-3-21_21-39-40.png


Step 3: Copy the above data to excel can calculate points per hour per logical CPU.See below for the WUs listed above. Here the average points per hour per CPU is about 34.24. I'm running E5-2695v2 under ubuntu 16.04 with 22 cores. 2 cores reserve for GPU crunching. So for SCC project, my average BOINC PPD is 34.24 pts/hr/CPU x 22 CPUs x 24 hours/day = 18,078 PPD.
upload_2017-3-21_21-50-52.png
 
VINA based apps in WCG perform much better in Linux so it would be good to narrow work to one app. I can lose like 10k BOINC RAC just be going from Zika to MCM as the latter is not a VINA app. PPD is supposed to be pretty similar between apps so I would guess it would be based off run times in Windows.
 
Progress so far.

Statistics Date
white.gif
Total Run Time (y:d:h:m:s)
white.gif
Points Generated
white.gif
Results Returned
03/22/2017
white.gif
0:011:15:55:22
white.gif
80,122
white.gif
142
03/21/2017
white.gif
0:010:14:07:19
white.gif
59,522
white.gif
167
03/20/2017
white.gif
0:000:07:20:16
white.gif
1,599
white.gif
6

Moved my GTX 1080 into the build today but it won't get any WUs until midnight tonight.
 
You must log in or register to reply here.
Back
Top