Author Topic: Threadripper Builds  (Read 52475 times)

2018-10-06, 14:00:27
Reply #135

fraine7

  • Active Users
  • **
  • Posts: 74
    • View Profile
Full specs for the Threadripper build:

CPU - 2990WX
GPU - Nvidia GTX1080Ti
MOBO - MSI X399 Sli-Plus (MS 7B09)
RAM - Corsair Vengeance CMU64GX4M4A2666C16 (64GB)
PSU - EVGA 1000 Watt GQ Gold Hybrid Modular
HD -  Samsung SSD 970 EVO 500GB (NvME)
Cooler - Thermaltake Riing 3.0 240mm AIO

Temps never really exceed 65c, monitored with CPUID HWmonitor

I haven't touched PBO (only just searched it online to see what it was) Is this something I should be looking to change?

Thanks
« Last Edit: 2018-10-06, 14:04:09 by fraine7 »

2018-10-06, 14:11:50
Reply #136

Juraj Talcik

  • Moderator
  • Active Users
  • ***
  • Posts: 3618
  • Tinkering away
    • View Profile
    • studio website
It looks in order, except for the cooler. Are you sure none of the cores exceeds 65 ? Your AIO cpu block doesn't cover the heatspreader, it definitely isn't 2990WX compatible cooler.

But I am not saying that is it.
talcikdemovicova.com  Website and blog
be.net/jurajtalcik   Our studio Behance portfolio
Instagram   Our studio Instagram, managed by Veronika

2018-10-06, 14:19:23
Reply #137

fraine7

  • Active Users
  • **
  • Posts: 74
    • View Profile
It may reach 67c on occasion but honestly the cooler seems to be doing a good job, even considering the lack of full coverage.

I am just looking at the PBO options in bios, if I set it to manual I was expecting to see some options for specifying a number but it doesn’t give me any. Do I set this number in the Ryzen master tool once enabled in Bios. I don’t actually know what this is or what it does to be perfectly honest.

Thanks

2018-10-06, 14:36:12
Reply #138

fraine7

  • Active Users
  • **
  • Posts: 74
    • View Profile
PPT, TDC and EDC are all maxed out at 250, 215 and 300 respectively. I’m in creator mode and these are the limits, I’m unable to type in a higher number - how did you get it to 350w?

2018-10-06, 14:42:28
Reply #139

Juraj Talcik

  • Moderator
  • Active Users
  • ***
  • Posts: 3618
  • Tinkering away
    • View Profile
    • studio website
I don't know if individual boards support different limits. In manual PBO mode in bios, I can specify anything and the MSI MEG limit is 700W. I didn't try Ryzen Master yet, it probably is helpful when trying different setups but ultimately it's better to set it up from boot.
talcikdemovicova.com  Website and blog
be.net/jurajtalcik   Our studio Behance portfolio
Instagram   Our studio Instagram, managed by Veronika

2018-10-06, 15:21:32
Reply #140

fraine7

  • Active Users
  • **
  • Posts: 74
    • View Profile
Apparently bios is the only way to do this, Ryzen Master is pretty useless to be honest, all values are capped at AMD’s recommendations (250w max for PPT)

My main concern is why I should be needing to change any of this stuff when all I am trying to do is use the chip at stock speeds to render. I haven’t even looked at overclocking as the stability is non-existent at 3ghz.

The lack of knowledge from my retailer regarding this chip and the x399 system is pretty frustrating, I might have to return both the cpu and motherboard and replace them with intel components but that would be a big performance loss and probably cost the same.

2018-10-06, 16:07:59
Reply #141

Juraj Talcik

  • Moderator
  • Active Users
  • ***
  • Posts: 3618
  • Tinkering away
    • View Profile
    • studio website
You shouldn't need to touch it at all. PBO is AMD's way of overclocking by simply unlocking the power limits. 250W is the upper limit of what 2990WX will require at stock. It would not be very stable without serious air flow to cool down the limited VRM phases of your board.


Can you check if both 12V leads (in upper left corner) are connected from PSU ? Just to rule out your system not getting enough juice.
Is your system stable if you run your memory at default clock without AXMP profile (AMXP OFF)?


MSI has very shit default bios settings. Make sure their GamingMode/OC Genie/etc..(it's called differently in every board) is turned SW(Software) and 0 (Turned OFF). It's in upper left corner in bios.
talcikdemovicova.com  Website and blog
be.net/jurajtalcik   Our studio Behance portfolio
Instagram   Our studio Instagram, managed by Veronika

2018-10-06, 16:27:00
Reply #142

fraine7

  • Active Users
  • **
  • Posts: 74
    • View Profile
Both CPU 12v leads are plugged in. System is absolutely fine until rendering, I could be using Photoshop or After Effects etc all day without issue but when I try set my animation off rendering it begins to cause instability, I have to manually switch off the psu as nothing else responds.

XMP has been disabled since the day I installed everything, I was hoping to enable it once everything seemed stable but I haven’t reached that point yet.

I’ve disabled the OC genie stuff too, and just as an extra step I’ve switched the physical switch to 0 on my mobo as well, although it is currently in SW (bios) mode.

Windows isn’t reading any errors or giving much useful info out in the reliability manager or event logs.

I’m thinking it’s a bad motherboard but could just as well be CPU.

A few years ago I had issues with the Asus Suite III and my old 5960x, the asus utility was trying to do things which Windows wasn't allowing (permissions etc) and it caused BSOD many times. I thought maybe this was a similar issue but even after uninstalling any kind of AMD utility the problem persists.
« Last Edit: 2018-10-06, 16:31:07 by fraine7 »

2018-10-06, 16:48:10
Reply #143

fraine7

  • Active Users
  • **
  • Posts: 74
    • View Profile
Currently running a stability test through Aida64.

CPU temp has not yet exceeded 68c at 100% load, 10 mins in and it is fixed at this value which suggests to me that the cooler is doing its job well enough.

Max power so far is 235w, this reaches 250w when rendering which could be an indicator of what is happening with the system hang/freeze/lockup etc. Could it be that the CPU is demanding more power but is being prevented from getting it by the bios settings, through which I am unable to exceed 250w.

Surely this can't be a Max/Corona issue, all the other stress tests I put this machine through suggest the build is super-stable but rendering is completely out of the question it seems.

Other thoughts I've considered, when rendering my animation, the CPU will obviously be at 100% when rendering but will then drop substantially whilst calculating geometry/scene parsing. Could there be an issue with the CPU picking itself up once the cores have relaxed for a moment?

2018-10-06, 17:00:39
Reply #144

Juraj Talcik

  • Moderator
  • Active Users
  • ***
  • Posts: 3618
  • Tinkering away
    • View Profile
    • studio website
But how is the temperature spread ? The cooler doesn't cover all of those four dies well so the cores on the outside can get disproportionally hotter. It's not unusual to get difference of 20C between different cores. That is basically the deal with my i9 which has small heatspreader, but the difference can be even bigger with the massive heatspreader of TR4.

I would run Prime95 which has AVX, it better simulates Corona's load.


Could there be an issue with the CPU picking itself up once the cores have relaxed for a moment?

The spike ? I doubt that. I do find Corona's pre-rendering to be poorly using CPU's performance though and this is even worse for some reason on Threadripper. Like I mentioned above, I only get 30perc. usage during UHD calculations.
talcikdemovicova.com  Website and blog
be.net/jurajtalcik   Our studio Behance portfolio
Instagram   Our studio Instagram, managed by Veronika

2018-10-06, 17:03:14
Reply #145

fraine7

  • Active Users
  • **
  • Posts: 74
    • View Profile
I will give Prime95 a run now, thanks.

I've attached the Aida64 test result in case there's anything which catches your eye - it was only a 20 minute test but this exceeds the amount of time I can render for without issue.

2018-10-06, 17:23:45
Reply #146

fraine7

  • Active Users
  • **
  • Posts: 74
    • View Profile
It seems neither Aida64 or Prime95 (currently still running) will push things to the 250w limit (see attached) - it only seems to go 90% of the way there, around 237w.

Corona seems to push this up to 250w immediately, I have not been able to crash the system running any stress test so far, not once.

Really confused as to what might be causing the rendering issues, I am currently on a 3.0 daily build and will revert back to 2.0 for further testing but I can't see that being the cause - worth a check though whilst I'm scratching my head for other ideas.

Thanks for your input so far Juraj, really appreciate it ;)

2018-10-06, 17:32:18
Reply #147

Juraj Talcik

  • Moderator
  • Active Users
  • ***
  • Posts: 3618
  • Tinkering away
    • View Profile
    • studio website
Hmm, I really don't know. So Corona does shoot to 250W, but that itself doesn't crash it if I understand, it's only after certain while (10 +/- minutes) when it does ? That would rule out power limits and point more towards cooling.

I can feel the frustration, when I've setup our 2990WX which we really needed at the time and it didn't even work with the public bios I was close to insanity. Every wasted day counted, 5k euro spent and shit is not working. If the private beta bios didn't solve my issues I was prepared to start with parts swap (first motherboard).


Hope this will be solved for you. One more try, maybe contact them through website ticket as well, there might be beta bios likewise.
talcikdemovicova.com  Website and blog
be.net/jurajtalcik   Our studio Behance portfolio
Instagram   Our studio Instagram, managed by Veronika

2018-10-06, 17:47:01
Reply #148

fraine7

  • Active Users
  • **
  • Posts: 74
    • View Profile
I think that's the next logical step, contacting AMD and seeking their advice.

My retailer who I get all my parts from are happy to run Aida64 and Prime95 tests on the chip but as I've already done those many times I don't think I would gain anything. They're not going to render my animation for me and keep an eye on it whilst it fails, would be useful if they did but this could well happen with another 2990wx if they were to replace it.

I'll let you know how I get on in case it helps with other Corona users looking to build a 2nd gen Ryzen machine!

2018-10-06, 17:51:17
Reply #149

Juraj Talcik

  • Moderator
  • Active Users
  • ***
  • Posts: 3618
  • Tinkering away
    • View Profile
    • studio website
I would strongly suggest for builders of 2990WX to stick to one of the two only boards designed to support it from the ground, MSI MEG and Aorus Xtreme. These are still imperfect and with poor bios, but they still stand out among the others.

I wonder if (and if it's actually coming in January, at least the reveal) X599 will change this.
talcikdemovicova.com  Website and blog
be.net/jurajtalcik   Our studio Behance portfolio
Instagram   Our studio Instagram, managed by Veronika