Ryzne 1700 + PReim95 V29.4B7 crashes

[SvenBent]

messing around with overclocking I am for the first time in many machines build a bit baffled.

If I run the newest prime95 v29.4B7 blend it will crash within a few minutes
at no overclockins set in bios (3.2ghz)

However its stable in
6 hours of PRime 95 v28.5 blend (However running bulldozer core)
7 hours of IBT very high
6 hours of Aida CPU/FPU/cahe/ram
11 hours of aida64 CPU/FPU


I have now clocked it up for 3.4ghz ad are running aida64 CPU/FPU test with no issues but only 3 hours so far.
So either my ryzen system is unstable at near stock speed, or there is something wrong with prime95 v29.4b7


Is there anyone with a ryzen system that can confirmd prime95 29.4b7 in blend mode are stable for a couple of hours?
or confirmed there is an issue with this specific prime version in ryzen?

 

[sirmonkey1985]

probably just a bad build.. try 29.4b5. that's the one i've been using.

 

[JustReason]

mesisng aorudn with overlcocking i am for the first time in many macinhes build a bit baffled.

i if run the newst prime95 v29.4B7 blend it will instanly crash within a few minuts
however its stable in

6 hours of PRime 95 v28.5 blend (However running bulldozer core)
7 hours of IBT very high
6 hours of Aida CPU/FPU/cahe/ram
11hours of aidag64 CPU/FPU

and this it just at 3.2ghz

I have now clocked it up for 3.4ghz ad are running aida64 CPU/FPU test with no issues but onmyl 3 hours so far.


So either my ryzem system is unstable at near stock speed. or there is something wrong with prime95 v29.4b7


Is there anyone with a ryzen system that can confirmed prime95 29.4b7 in blend mode are stable for a couple of hours ?
or confirmed there is an issue with this specific prime version in ryzen ?

Prime has been a pain in the ass for the last few years for both Intel and AMD.

 

[SvenBent]

probably just a bad build.. try 29.4b5. that's the one i've been using.

Do you have a link for it ?

Prime has been a pain in the ass for the last few years for both Intel and AMD.

Sorry but iI ant agree on this. All my previous machines (Both AMD and intel) have always been stable in prime95. Heck my wifes current computer ran prime95 28.5 blend mode for 36+ hours.
A lot of people seem to blame prime95 for doings its job, instead of dealing with the realizations that the system is unstable.

However in this case we are talking prime unstable at stock speed and only a specific version and everything else seems stable so far at overclocke


11hours aida 64 CPU/FPU at 3.4ghz currently running IBT very high

 

[sirmonkey1985]

overclock.net has it but they're upgrading their forums. if it's stable in aida and ibt i'd call it good.

justreasons not wrong though, ever since they added the newer avx crap p95 has had issues.

 

[SvenBent]

overclock.net has it but they're upgrading their forums. if it's stable in aida and ibt i'd call it good.

justreasons not wrong though, ever since they added the newer avx crap p95 has had issues.

I again i must disagree on this never had issues runnig AX speed but see plenty of people with excused of why this and that when basically its just an unstable system (albeit very tiny)
AVX is part of the specs. If your CPU can't run it something is wrong and need to be adjusted ( or accept the instability).
aida64 is for sure weaker than prime95 which is why i dont like accepting it as a tresshold for anything in regards to stability in general. saying ADaida is ok so ignore prime resutls is the same as saying it stable in word so ignore aidag. just scaled down. the system is unstable if it can't do its calculations correct.
IBT is slightly diffrent at normalt its weaker than prime95 but at very high is seems comparable if not slightly stronger.


Anyway I found the 29.4b5 but that one runs AVX core on my system. 29.4b7 runs FMA3. at least is an upgrade from 28.5 that ran bulldozer core.
I just really want to make sure if its a prime issue with FMA3 core on Ryzen or indeed the CPU doing miscalculations.


No to mention this is at near stock if nto at stocks speed (no OC in bios) if its this unstable at this speed there is no way i will be able to overclock it and keept it stable.

Right now it just feels like a downgrade from my old OC'ed 3770k

 

[Araxie]

AVX it's part of the specs but not at the unrealistic levels of load by prime95 or aida64, this latest which you are partially using wrong, aida64 is not weaker than prime95, you are just running CPU+FPU when you should be running it at FPU only for maximum stress levels (higher than p95 smallFFT), you will not find anything in the real world ever that put AVX stress load at p95/aida levels.

And there's where the other "thing" play a role, as you know because you always say, there's a difference between CPU stability and system stability, p95 blend it's more system stability than CPU stability, if it pass with small/large fft but fails with blend then it's a matter of system stability, depending on your motherboard at stock speed it may just be applying a poor voltage regulation or weak VRM phase regulation at stock but it may run better when OC'd, some motherboard run "optimized" settings when overclocking and even may force to disable powersaving features when overclocking which also may be adding stability to the mix as Ryzen require a very very very tight power control at stock to run under specificied TDP which for your CPU it is only 65W, then again, some motherboard when overclocking just bypass this and add a lot extra TDP headroom by itself, you will have just to mess with the BIOS and learn everything about it.

My MSI B350 Tomahawk and my 1700x even just adding one CPU multiplier, it is considered overclock and disable most power saving features and that piss me off because it like to have Cool'N Quiet, but also it increase TDP from 95W to 125W and VRM to optimized values with higher current support and faster answer by the circuitry, but my Asus C6H doens't behave the same, it allow manual and granulated control for everything. and there's where I think your instability is capping the Ryzen 1700 at 65W it's just plain bad for such stress test as prime95 blend.

 

[JustReason]

overclock.net has it but they're upgrading their forums. if it's stable in aida and ibt i'd call it good.

justreasons not wrong though, ever since they added the newer avx crap p95 has had issues.

Thx. I was referring mostly to the Intel issue, think Skylake, Where it was either causing overvolting or too much heat, so there was a warning to avoid Prime till an update was made. On the FXs from AMD it was mostly bad as the code was not kind to the shared FPU and always caused that one core to fail early. My 8350 passed everything else no prob but Prime just wasn't having any of it, even at stock. 5 years and had no one single crash or hardware error, so Prime is not a go to for me at all. I kinda prefer IBT for the lower time requirements.

 

[SvenBent]

3+ hours so far in prime29.4b5 AVX load. it looks more and morel ike it was just prime 29.4b7 that was just bad

 

[Mega6]

I just use IBT and I don't run it for days. High setting usually does the trick.

 

[SvenBent]

I just use IBT and I don't run it for days. High setting usually does the trick.

Since IBt is jsut a frontend for Linpack which is a intel math library. there is good reaosn to beleive its not optimizef for AMD architecture and thereby does not maintains its strenght. which is why i am a little hesittent to trust IBT
besides IBT is also using a long outdate linpack i ussually use linpack directly myself.


anyway prime 29.4b5 have been running close to 15hours now. So I am pretty sure the erros seen with 29.4b7 is an software error in prime.

anyone know a stressfull FMA based test?