BIOS Updates Coming For Skylake Freeze Bug

H

HardOCP News

[H] News
Joined
Dec 31, 1969
Messages
0
According to this post on the Intel community forums, a bug causing Skylake processors to freeze while running applications like Prime 95 has been fixed. Those of you running Skylake processors will have to update your BIOS once the fix is available from your motherboard manufacturer.

Intel has identified an issue that potentially affects the 6th Gen Intel® Core™ family of products. This issue only occurs under certain complex workload conditions, like those that may be encountered when running applications like Prime95. In those cases, the processor may hang or cause unpredictable system behavior. Intel has identified and released a fix and is working with external business partners to get the fix deployed through BIOS.
 
Prime95 has been one of my go to tools for stability testing.
This would have driven me crazy.
 
Shows the power of prim95 for stress testing. :D

That and linpack has been my goto CPU stresstools for ages
They also come out far ahead of any other stress tools I've tested ( Superpi. wprime. CPU-burnin CoreDamage S&M 7-zip Passmark-burnin. hot-cpu etc)
 
Noghri2 said:
I know its not the same, but I remember when original Pentiums had flaw in them :

https://en.wikipedia.org/wiki/Pentium_FDIV_bug
Click to expand...
lol yeah, that's why Intel started making microcode/state updates loadable via firmware. Now instead of making landfill material out of buggy processors, it can fix most problems. The FDIV bug Pentiums had a problem with a ROM table built in the chip, which contained some incorrect values.

Apparently this particular Skylake bug is in AVX, and it's pretty difficult to trigger. Intel seems to have figured out a way to fix it in about a month.
 
In the case of the FDIV issue, you could work around it in software, but the workaround slowed the offending operation down so much that you might as well be running a 386.
 
GiGaBiTe said:
In the case of the FDIV issue, you could work around it in software, but the workaround slowed the offending operation down so much that you might as well be running a 386.
Click to expand...
There were various workarounds for the bug. The official workaround code Intel provided (check operand bit patterns and apply specific fix term if found) slowed down FDIV heavy code by around 20% IIRC.

Another way was to trap the FDIV instruction and do software emulation of affected division, which would be much slower than a 387 coprocessor, but much faster than a 386 CPU doing the same thing. That was kind of pointless since better solutions were available via recompiling (after compilers were updated) or using the official workaround manually.
 
I thought for a while now that testing with Prime95 wasn't recommended anymore. My 2 X z170 systems have been tested with Asus Real Bench and I haven't seen any stuttering or anything like that.
 
I was kind of annoyed that SGX is now enabled on chips manufactured after October. Not that I have any planned use for the SGX instructions as of yet, but an "updated" Skylake part after the fact smacks of a half baked process - and I only bought my 6700K in November.

Maybe their retooling/fixing/updating of the Skylake core has something to do with the shortages.
 
SvenBent said:
Shows the power of prim95 for stress testing. :D

That and linpack has been my goto CPU stresstools for ages
They also come out far ahead of any other stress tools I've tested ( Superpi. wprime. CPU-burnin CoreDamage S&M 7-zip Passmark-burnin. hot-cpu etc)
Click to expand...

This problem only occurred with certain versions of Prime95.
 
Could you imagine being someone pulling their hair out about not being able to get a stable OC :D
 
Good thing I was using the Intel stress tool with my latest Skylake build. I had also heard prime95 was not the tool to use anymore either.
 
bigdogchris said:
Could you imagine being someone pulling their hair out about not being able to get a stable OC :D
Click to expand...

Monkey God said:
Good thing I was using the Intel stress tool with my latest Skylake build. I had also heard prime95 was not the tool to use anymore either.
Click to expand...

Intel's Burn in Test often puts unrealistic workloads on CPUs. I'm not saying you shouldn't use it, but a failure in that tool doesn't necessarily mean your overclock is bad. You will often see temperatures in the Intel test which you won't be able to achieve any other way. As for Prime95, we've often seen things like Handbrake encoding crash Skylake systems with overclocks that pass as Prime95 stable.
 
Dan_D said:
Intel's Burn in Test often puts unrealistic workloads on CPUs. I'm not saying you shouldn't use it, but a failure in that tool doesn't necessarily mean your overclock is bad. You will often see temperatures in the Intel test which you won't be able to achieve any other way. As for Prime95, we've often seen things like Handbrake encoding crash Skylake systems with overclocks that pass as Prime95 stable.
Click to expand...

Makes me feel even more justified to hold on to IB and wait for Kaby Lake or Cannonlake...
 
Dan_D said:
~
As for Prime95, we've often seen things like Handbrake encoding crash Skylake systems with overclocks that pass as Prime95 stable.
Click to expand...

Prime95 28.7? That I'd love to see a video of and not just take your word for it.
 
Shit it used to be..either you cpu was bullet proof with EVERYTHING or it wasn't truly bullet proof stable....not the new cpus these days.....Its hard to explain but i feel like these older cpus are much more durable? if thats the right word...it just doesn't make me want to run out and upgrade if ya got to be careful not to even use prime95 or this or that lol..fuck when do the real cpus come out again?
 
primetime said:
Shit it used to be..either you cpu was bullet proof with EVERYTHING or it wasn't truly bullet proof stable....not the new cpus these days.....Its hard to explain but i feel like these older cpus are much more durable? if thats the right word...it just doesn't make me want to run out and upgrade if ya got to be careful not to even use prime95 or this or that lol..fuck when do the real cpus come out again?
Click to expand...

People are just using selective confirmation. When they get the stability they want in most apps, one other app that causes a crash is considered the outlier and "not a real test". Like how Haswell chips were running too hot or crashing in Prime 95 because of the app using AVX instructions and the chips automatically boost voltage when running AVX. Or when the Q6600 launched and "everyone" could get their PC stable at 3.6 with some rather limited testing. Then summer came along and people had to tone down the overclocks because those systems started crashing.
 
You must log in or register to reply here.
Back
Top