Overclock.net › Forums › AMD › AMD CPUs › AMD Ryzen Threadripper Owners Club - 1950X | 1920X | 1900X
New Posts  All Forums:Forum Nav:

AMD Ryzen Threadripper Owners Club - 1950X | 1920X | 1900X - Page 122

post #1211 of 2389
Quote:
Originally Posted by Particle View Post

I did some tests with firmware 1.50. It doesn't seem to be a hardware issue.

Stock settings: Blender crash
Stock settings, LLC set higher: Blender crash
Stock settings, turbo disabled: Blender crash
Stock settings, turbo disabled, SMT disabled: Blender crash
Stock settings, turbo disabled, SMT disabled, max multiplier set to 22 (2.20 GHz): Blender crash

It's always about 3-5 seconds into the tile based renderer part where it happens.

i do not know if it related but i was getting crashes and slow times on blender so i grabbed whatever was the latest, uninstalled the old one and simply ran blender from the folder without installing and no more issues...again not sure if its the same thing.
blender-2.79-rc2-windows64

is the one i grabbed
post #1212 of 2389
Warning: Spoiler! (Click to show)
Quote:
Originally Posted by Simmons572 View Post

I'm ready to be added to the club!



This is the current state of my build. Once I get the XSPC waterblock, I will clean this up quite a bit. smile.gif

Specs:
  • Threadripper 1900x
  • Asus X399-A Prime
  • G.Skill - Trident Z 32GB (4 x 8GB) DDR4-3200
  • Asus 980 ti Matrix


I also installed the reservoir in the rear of the case, in preparation for the loop. More pics in my build log if you all are interested!

noice
now go run the kraytace bench for us (links in the thread somewhere smile.gif )seems there is only a 1900x with dual channel ram i,m dying to see what benefits quad gives...if any wink.gif
Quote:
Originally Posted by Bm514 View Post

I find that Prime 95 Is a bit unrealistic of a load I use aida64 Cpu Fpu and Cache I get a Max Of 66 after 36 Hours

i find none of them by themselves are any real good especially artificial ones what chew does seems the most vicious method i basically start with cb15 cpuz bench then real bench 2.54 and move up fro there with the rendering benches i generally leave gooseberry till last as that's nasty smile.gif

if that all goes well i will fire up prime 95 and the other waterboarding tests.
post #1213 of 2389
Quote:
Originally Posted by Bm514 View Post

I find that Prime 95 Is a bit unrealistic of a load I use aida64 Cpu Fpu and Cache I get a Max Of 66 after 36 Hours

It depends what you use your computer for. I use mine to solve some pretty complex stuff, and even though my chip is aida64 stable, it crashes in the solving software I use. So I rely on P95 to find my stable point
post #1214 of 2389
We are using aegis photoscan professional. It uses gpu cpu and mem and thrashes it hard non stop. A project can take days.

Thus we stress test mission critical.

2 days in then a crash is 2 days of lost work.

Like i said. For those that bought TR with the full intent to use it as it was intended...I do not realistically see 1.4v and higher as a 24/7 voltage

We have crashed trying p states and letting cpu go up down...which it does a whole hell of alot of running this.
Solution just run a fixxed OC.

We have crashed due to hitting a mem limit. Simple solution add 64gb more.

Fyi the memory needed for this project costs more than TR.

Anyway to each his own we all play our roles. Mine is tuning mission critical systems and I need to roll with what I know is reliable not what I think is reliable.

Fyi @ 1.2v @ 3.8 i am seeing 5c less than prime 95 and only 3c from AMD specced safe range wink.gif

Im on a swiftech h220x2 with a apogeee drive x2.

Fans are not stock...quite loud and annoying.

Copper rad vs alum on crap aio...plenty of pump power and a very large surface area on coldplate over dies.

We can improve temps...i am willing to bet with an aio we can not however.
Edited by chew* - 9/22/17 at 9:09pm
post #1215 of 2389
Quote:
Originally Posted by chew* View Post

We are using aegis photoscan professional. It uses gpu cpu and mem and thrashes it hard non stop. A project can take days.

Thus we stress test mission critical.

2 days in then a crash is 2 days of lost work.

Like i said. For those that bought TR with the full intent to use it as it was intended...I do not realistically see 1.4v and higher as a 24/7 voltage.


+1 I dont much like having lost a full days work of solving myself. Especially since I run my computer hard almost 24/7. On top of that, having 1.4v+ and 70c+ constantly will likely result in degradation of the chip as time goes on. Reliability is key for me.

For people who game or are just enthusiasts, I guess you can run it as high as you want for bragging rights, since gaming will never stress your computer hard. At the end of the day, failing p95 probably doesnt mean much if you are not a content producer using cpu intensive software.
post #1216 of 2389
You should probably be running ECC then.
post #1217 of 2389
Quote:
Originally Posted by L33tBastard View Post

You should probably be running ECC then.

Probably should....but ecc is A. More expensive.

B. Specced for 2400.

C much slower and ryzen needs the higher speed to hit the performance we want.

If your anal at tuning and stress testing ecc is not needed...just run a stable tune...

This is the benefit of TR. Getting server grade performance without server grade parts that have some flexibility with clocks and tuning.
Edited by chew* - 9/22/17 at 9:16pm
post #1218 of 2389
Understandable, yet you can overclock ECC ram on this platform like some already do while retaining it's function of error correction.

But yeah, you cannot really escape it's price, specially now days... crazy pricing.
post #1219 of 2389
Quote:
Originally Posted by L33tBastard View Post

Understandable, yet you can overclock ECC ram on this platform like some already do while retaining it's function of error correction.

But yeah, you cannot really escape it's price, specially now days... crazy pricing.

Yah i went and looked at prices and did some research.

I was like wow...might as well just stick with the b die 3400 16-16-16 bin...at least I know it works and what expectations to have from it.

The saving grace is we are in 64gb @ around $500 range due to my bargain hunting.

Even if we pay full price we are ahead of the game on another 64gb.

Thx to gupsterg i am another $200 ahead of game on waterblocks for fury x's but thats just my personal system.

This system will be running quadros or at least i think thats what he is loading in it.

We are just using mine to tax psu, board, pci lanes for now.

Got about another week to get it dialed in perfect before it goes from demo use to full 24/7 use.
Edited by chew* - 9/22/17 at 9:46pm
post #1220 of 2389
Quote:
Originally Posted by chew* View Post

Hard ocp did a write up on a byski and xspc block on xspcs flagship double thick 480 rad.

Long story short 75-76c @ 1.4xx peak temps in prime 95

So basically if not 68c or less not a voltage I would run for 24/7 mission critical.

Thx for posting by the way i was guesstimating 3066 for 128g but unsure.

We already ran out of mem with 64g on a project which btw eats all 32 threads and is seeing temps like prime after hours of running.

Basically if you use the system the way it was intended...prime is realistic.

Which Prime test do you run and how long?
New Posts  All Forums:Forum Nav:
  Return Home
  Back to Forum: AMD CPUs
Overclock.net › Forums › AMD › AMD CPUs › AMD Ryzen Threadripper Owners Club - 1950X | 1920X | 1900X