Overclock.net - An Overclocking Community - Reply to Topic
Thread: Ryzen Threadripper 2990WX, 2970WX, 2950X, 2920X owners club Reply to Thread
Title:
Message:

Register Now

In order to be able to post messages on the Overclock.net - An Overclocking Community forums, you must first register.
Please enter your desired user name, your email address and other required details in the form below.
User Name:
If you do not want to register, fill this field only and the name will be used as user name for your post.
Password
Please enter a password for your user account. Note that passwords are case-sensitive.
Password:
Confirm Password:
Email Address
Please enter a valid email address for yourself.
Email Address:

Log-in


  Additional Options
Miscellaneous Options

  Topic Review (Newest First)
12-30-2019 05:52 PM
cdawall Putsing around with some good memory today. I am curious if I can get a 4 dimm load fully stable at this speed, but these Bdie sure make a world of difference.

12-21-2019 08:11 AM
cdawall
Quote: Originally Posted by bogdi1988 View Post
I had my MSI Creator board fail a year after I got it. Machine would boot but in a low frequency/voltage state at 600MHz per core with .75Vcore. I would have to manually set up the voltage in bios for it to post normally. Definitely some issues with the VRM/electrical distribution of the board.
Well I have not beaten this poor board up that bad yet, but I am sure my 24/7 clock is not doing it any favors. This runs cooler and at lower voltage over letting PBO do it's thing.

12-17-2019 06:13 PM
ugotd8 Hi all. This is some data from the https://opendata.blender.org/ site which is compiling blender benchmark statistics when uploaded. They have a 'top50 cpu' list and some other data but they also have 'raw data' available for download. It's a JSON file with every entry ever submitted. I wrote some python code to pull out information on the recent AMD HEDT and high core count mainstream procs from the raw data to help me decide which processor to upgrade to. (footnote: I know the 2990WX did better in linux but I don't run linux so this list is purely windows results)

Blender Opendata 'QUICK' benchmark data for recent AMD processors on Windows

AMD Ryzen Threadripper 3970X (4 entries): fastest: 253 average: 261 (55.15% speedup)
AMD Ryzen Threadripper 3960X (8 entries): fastest: 310 average: 349 (40.03% speedup)
AMD Ryzen Threadripper 2990WX (63 entries): fastest: 296 average: 373 (35.78% speedup)
AMD Ryzen Threadripper 2970WX (18 entries): fastest: 383 average: 471 (18.93% speedup)
AMD Ryzen 9 3950X (23 entries): fastest: 444 average: 488 (16.04% speedup)
AMD Ryzen Threadripper 2950X (50 entries): fastest: 573 average: 636 (-9.39% slower)
AMD Ryzen 9 3900X (295 entries): fastest: 593 average: 665 (-14.53% slower)
AMD Ryzen Threadripper 1950X (102 entries): fastest: 538 average: 673 (-15.74% slower)

The speedup percentage above is relative to my 'old' 1950X overclocked to 3.9Ghz all core which scored 584 on this test. The math is done against the average,not the fastest for each CPU. All scores in seconds, lower is obviously faster.

Blender Opendata 'COMPLETE' benchmark data for recent AMD processors on Windows

AMD Ryzen Threadripper 3970X (1 entries): fastest: 959 average: 959 (55.35% speedup)
AMD Ryzen Threadripper 3960X (2 entries): fastest: 1207 average: 1219 (43.23% speedup)
AMD Ryzen Threadripper 2990WX (19 entries): fastest: 1260 average: 1449 (32.51% speedup)
AMD Ryzen Threadripper 2970WX (6 entries): fastest: 1580 average: 1757 (18.17% speedup)
AMD Ryzen 9 3950X (6 entries): fastest: 1620 average: 1761 (18.02% speedup)
AMD Ryzen Threadripper 2950X (12 entries): fastest: 2154 average: 2304 (-7.28% slower)
AMD Ryzen 9 3900X (71 entries): fastest: 2135 average: 2382 (-10.91% slower)
AMD Ryzen Threadripper 1950X (27 entries): fastest: 1979 average: 2443 (-13.76% slower)

The speedup percentage above is relative to my 'old' 1950X overclocked to 3.9Ghz all core which scored 2148 on this test. The math is done against the average, not the fastest for each CPU. All scores in seconds, lower is obviously faster.

I think the most interesting thing from this data is how the 3950X is virtually tied with the 2970WX of last gen.

Granted these are small sample sizes, the 3000 procs haven't been out that long and not everyone is submitting complete runs to blender opendata site.

Hope some find this information useful, it's not the whole picture but how these do in Blender is at the top of my list.
12-15-2019 11:58 PM
bogdi1988
Quote: Originally Posted by cdawall View Post
So it has been a while since I posted in here, figured I would drop in and give some insight for those looking at these processors and the different boards as they go EOL for the 3xx0 series releases. We are at like 15 months of ownership and I am seeing the wear on the VRM section without a doubt. The BIOS updates have brought my CPU voltage way down however so it somewhat balances out in the long run. Without a doubt I cannot recommend anything with the standard 60A and 8 phase design for overclocking a 2990WX. AVX load and 4.1ghz easily overloads the VRM section. Without at minimum active cooling the VRM section cooler becomes too hot to touch. On an interesting note you can actually hear the VRM section go into low power mode when it flips between 4.1ghz and 55x mhz as safeties built into the motherboard hit. With the VRM section temperature dropped with cold air coming across it (40F actively cooled) I can get the chip a lot higher clock wise with the current BIOS. A proper VRM section could have the better binned chips to at least 4.3 all cores.

With some very heavy CPU tweaking this is what I found to be the max 24/7 stable setting I could pull of assuming I could keep the VRM section cool. Very good performing combo when combined with b-die. There is a lot more ability left in this ram I have not had a chance to even push it yet. I could easily have another 300-500 points locked away in that.



I do hope to see a TRX40 MATX board to replace my current setup when the time comes, but am not holding my breath.
I had my MSI Creator board fail a year after I got it. Machine would boot but in a low frequency/voltage state at 600MHz per core with .75Vcore. I would have to manually set up the voltage in bios for it to post normally. Definitely some issues with the VRM/electrical distribution of the board.
12-08-2019 10:43 AM
cdawall So it has been a while since I posted in here, figured I would drop in and give some insight for those looking at these processors and the different boards as they go EOL for the 3xx0 series releases. We are at like 15 months of ownership and I am seeing the wear on the VRM section without a doubt. The BIOS updates have brought my CPU voltage way down however so it somewhat balances out in the long run. Without a doubt I cannot recommend anything with the standard 60A and 8 phase design for overclocking a 2990WX. AVX load and 4.1ghz easily overloads the VRM section. Without at minimum active cooling the VRM section cooler becomes too hot to touch. On an interesting note you can actually hear the VRM section go into low power mode when it flips between 4.1ghz and 55x mhz as safeties built into the motherboard hit. With the VRM section temperature dropped with cold air coming across it (40F actively cooled) I can get the chip a lot higher clock wise with the current BIOS. A proper VRM section could have the better binned chips to at least 4.3 all cores.

With some very heavy CPU tweaking this is what I found to be the max 24/7 stable setting I could pull of assuming I could keep the VRM section cool. Very good performing combo when combined with b-die. There is a lot more ability left in this ram I have not had a chance to even push it yet. I could easily have another 300-500 points locked away in that.



I do hope to see a TRX40 MATX board to replace my current setup when the time comes, but am not holding my breath.
12-06-2019 04:10 AM
Redwoodz
Quote: Originally Posted by Daniel Ray View Post
Is it safe to update to most recent bios from a very old one? I am still on bios from around a year ago.
You have to follow your motherboards support instructions. You may have to install your bios update in steps.
12-06-2019 01:24 AM
Daniel Ray Is it safe to update to most recent bios from a very old one? I am still on bios from around a year ago.
11-29-2019 07:41 PM
J7SC ...Silicone lottery is always omnipresent...but really, I probably learned more with the slouchy pig oc chips (my old nemesis >> 5960x comes to mind) than the smooth operators, like the current 2950x. Also, you're right about the 'weird' RAM multipliers w/ TR2...3400 ok, 3466 ok, 3600ok, but 3533 not ok at any setting whatsoever ...of course, it could be all in the bios...
11-29-2019 04:58 PM
Bart
Quote: Originally Posted by J7SC View Post
I didn't try the 1usmus tool until version 1.6, well after I had already established a stable RAM, and by that time, the tool was quite mature and very helpful, if just to confirm things. And while I have done a lot of oc'ing, including with LN2, I also reached the age whereby spending three weeks on troubleshooting one particular obscure RAM timing parameter doesn't really makes sense, especially as RAM oc'ing still involves a certain amount of 'voodoo'...

Most reviews I have read seem to suggest that tight 3200 RAM is the sweet-spot. If/when you can reach that with your 16GB modules for 64GB on 2950X, it's eggnog time !
Nice, I'll keep that in mind. Silicon lottery is still a factor too, and AMD CPUs can be downright weird. My original 1st gen Ryzen 1700 wouldn't run RAM at 3200 no matter what, tried multiple kits, but for some reason 3466 was just fine (discovered that purely by accident LOL). Currently I have the ram manually set to 3200, and I manually input every single number from the DOCP 3600 timings, just to play safe. Once I get the loop set up properly, I'll start with looking at the timings.
11-29-2019 04:25 PM
J7SC
Quote: Originally Posted by Bart View Post
I've tried it, and honestly didn't like it. It spit out weird timings I knew my RAM couldn't do, so I wouldn't trust it. But it might be worth revisiting, since I know that guy works hard on that tool, so maybe the old version I was using was buggy. I really don't enjoy OCing like I used to, so I tend to avoid stuff like this until I have to. Like now. These days as an old guy, stability is priority #1. Squeezing every last bit of performance out of it isn't worth the hassle, unless you really enjoy the tinkering part of it. Thanks for the heads up, I will take a look at the latest version of that tool.

I didn't try the 1usmus tool until version 1.6, well after I had already established a stable RAM, and by that time, the tool was quite mature and very helpful, if just to confirm things. And while I have done a lot of oc'ing, including with LN2, I also reached the age whereby spending three weeks on troubleshooting one particular obscure RAM timing parameter doesn't really makes sense, especially as RAM oc'ing still involves a certain amount of 'voodoo'...

Most reviews I have read seem to suggest that tight 3200 RAM is the sweet-spot. If/when you can reach that with your 16GB modules for 64GB on 2950X, it's eggnog time !
This thread has more than 10 replies. Click here to review the whole thread.

Posting Rules  
You may post new threads
You may post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off