Author Topic: Threadripper Builds  (Read 157288 times)

2020-01-17, 18:55:49
Reply #840

dfcorona

  • Active Users
  • **
  • Posts: 150
    • View Profile
I think I'm looking to upgrade to the 3970x, seems to make more sense than the 3990x since it's about 60%+ the performance but faster with single core performance which will still keep workflow snappy.

Man my 1950x is running a 7159 in Cinebench R20 but only a 1min15sec in Corona.  wonder whats wrong. Running G.SKILL TridentZ Series 64GB (8 x 8GB) 288-Pin DDR4 SDRAM DDR4 3200, so don't think it's the ram.

I attached Temps, how do they look? can that be the issue?
« Last Edit: 2020-01-17, 19:04:32 by dfcorona »

2020-01-17, 20:44:27
Reply #841

JoeVallard

  • Active Users
  • **
  • Posts: 129
    • View Profile
    • Joe Vallard
Man my 1950x is running a 7159 in Cinebench R20 but only a 1min15sec in Corona.  wonder whats wrong. Running G.SKILL TridentZ Series 64GB (8 x 8GB) 288-Pin DDR4 SDRAM DDR4 3200, so don't think it's the ram.
I attached Temps, how do they look? can that be the issue?

Whats your issue? Cause those scores look like your running the 1950x with no overclock is all. Other than that looks normal to me. I have a 2950x OC'd @ 3.9Ghz, R20 = 7300 and Corona 1.3 = 1:07

Didn't order yet memory.

Did you ever decided which TR you were going with?

2020-01-17, 20:54:59
Reply #842

dfcorona

  • Active Users
  • **
  • Posts: 150
    • View Profile
Man my 1950x is running a 7159 in Cinebench R20 but only a 1min15sec in Corona.  wonder whats wrong. Running G.SKILL TridentZ Series 64GB (8 x 8GB) 288-Pin DDR4 SDRAM DDR4 3200, so don't think it's the ram.
I attached Temps, how do they look? can that be the issue?

Whats your issue? Cause those scores look like your running the 1950x with no overclock is all. Other than that looks normal to me. I have a 2950x OC'd @ 3.9Ghz, R20 = 7300 and Corona 1.3 = 1:07
Well I am running a 3.8ghz overclock, guess I expected lower time. Hard to tell by the "Freq. [GHz]" posted on results, they don't seem right.

2020-01-17, 21:25:00
Reply #843

Juraj Talcik

  • Moderator
  • Active Users
  • ***
  • Posts: 3947
  • Tinkering away
    • View Profile
    • studio website
Hard to tell by the "Freq. [GHz]" posted on results, they don't seem right.

They're indeed wrong. The benchmark can't reputably read clocks. Your temps are good, there is no throttling, it's even written there = ). 3.8GHz is only 200 MHz more than stock all-core boost. Also, Corona can apply AVX offset, are you sure it's actually running at 3.8 during the benchmark?

Did you ever decided which TR you were going with?

I am decided on 3990X unless the benchmarks reveal some kind of "catch" in performance (regardless of what due, Windows or architecture, or perhaps Corona).


talcikdemovicova.com  Website and blog
be.net/jurajtalcik   Our studio Behance portfolio
Instagram   Our studio Instagram, managed by Veronika

2020-01-17, 22:25:33
Reply #844

Michael Arch-Viz

  • Active Users
  • **
  • Posts: 65
    • View Profile
i found some interesting facts!
I connected the second cooler, same model and you know guys , while rendering my temp with two coolers is 68-70.
With only one cooler temp goes up to 81c. :)




2020-01-17, 23:00:04
Reply #845

diffuus

  • Active Users
  • **
  • Posts: 20
    • View Profile
Correct me if I am wrong, but is that the Y-cable NA-YC1 you're using to connect the two fans to the same header? That would reduce the fan speed to half.
Strange to see that 2 fans at half the speeds perform so much better than one at full speed.
Isn't there a spare header on the motherboard to connect both individual? Or is this done on purpose?

2020-01-17, 23:05:44
Reply #846

Michael Arch-Viz

  • Active Users
  • **
  • Posts: 65
    • View Profile
yes, you are right, i had this cable from my two years old noctua :)
i think this works better for me because my setup is open and i need some airflow which will take out hot air from CPU. i will make new test tomorrow with case and fans.

i have now on rendering


« Last Edit: 2020-01-17, 23:12:31 by Michael Arch-Viz »

2020-01-17, 23:17:50
Reply #847

diffuus

  • Active Users
  • **
  • Posts: 20
    • View Profile
That seems more like full speed. Then the Y-cable will have no influence on the speed. My mistake :-)

2020-01-18, 00:05:18
Reply #848

Michael Arch-Viz

  • Active Users
  • **
  • Posts: 65
    • View Profile
UPDATE

i opened the window and fresh air comes in the room, look how temp dropdown



2020-01-18, 02:22:57
Reply #849

lupaz

  • Active Users
  • **
  • Posts: 425
    • View Profile
Correct me if I am wrong, but is that the Y-cable NA-YC1 you're using to connect the two fans to the same header? That would reduce the fan speed to half.
Strange to see that 2 fans at half the speeds perform so much better than one at full speed.
Isn't there a spare header on the motherboard to connect both individual? Or is this done on purpose?

Even if the Y-cable reduces the speed by half (are you sure about this?), if the temp rises the speed keeps rising until the temperature goes down, right. I also have it with the Y-cable, and I can hear the fans speeding when rendering. I interpret that as it's working fine. No?

2020-01-18, 11:31:07
Reply #850

diffuus

  • Active Users
  • **
  • Posts: 20
    • View Profile
No, it was my mistake. The Y-cable doesn't reduce the fan speed.

2020-01-19, 14:55:51
Reply #851

michaltimko

  • Active Users
  • **
  • Posts: 286
  • Coronaut (c)
    • View Profile
I was testing push pull setup extensively and difference on 2990wx was like 1C at most.

Its hard to believe you are getting such huge difference. Its still single tower.
« Last Edit: 2020-01-19, 20:01:59 by michaltimko »
Coronaut!(c)2011

Supporting Corona in commercial projects since pre-alpha

2020-01-20, 11:13:09
Reply #852

psanitra

  • Active Users
  • **
  • Posts: 36
    • View Profile
    • NeEmotion.net
My Zenith II + 3970X came in today, will replace my dying x399 Aorus Extreme board with 2990wx. I went Zenith this time because i use lot of NVME drives,NVME raid 0 too, so dimm.2 slot is very handy.
Noemotion.net - www.noemotion.net
Peter Sanitra - www.psanitra.com
Noemotionhdrs.net - www.noemotionhdrs.net

2020-01-20, 11:48:09
Reply #853

Juraj Talcik

  • Moderator
  • Active Users
  • ***
  • Posts: 3947
  • Tinkering away
    • View Profile
    • studio website
Shiny and chrome = ). Dimm.2 is fantastic feature on ROG boards, I wish they would put two of them on like on Dominus board and make the board even wider.

The copper heatsink extension card from Aorus is like 1kg heavy. Too bad my fileserver is not pci4.0 so it won't bifurcate this one properly even for pci3.0 cards.

What do you use Raid 0 NVMe for?
talcikdemovicova.com  Website and blog
be.net/jurajtalcik   Our studio Behance portfolio
Instagram   Our studio Instagram, managed by Veronika

2020-01-20, 15:41:23
Reply #854

psanitra

  • Active Users
  • **
  • Posts: 36
    • View Profile
    • NeEmotion.net
What do you use Raid 0 NVMe for?

As main work drive. I do a lot of heavy Houdini sims and need both size and speed. Like on current job i do a 2-3 sim iterations a day,  roughly 1.5TB written, and maybe 2-4TB read back daily, Doing viewport previews, renders, converting volumes to vdbs, alembics, that kind of stuff, i can utilize both 32cores and NVME drives with running 1-8 deadlines slave on single workstation. Some times i can see disk the utilization peak over 50% on raid 0, running 2x2TB Samsung Evo. I wonder how long those drives will last...
Noemotion.net - www.noemotion.net
Peter Sanitra - www.psanitra.com
Noemotionhdrs.net - www.noemotionhdrs.net