What Does Prime Stress That LinX & OCCT Dont??

JonJ678 said:
Not a windows only chap then :D

Given me a few things to google there, cheers. Not my field of expertise at all.

I sympathise with linux being unstable. Much as I tend to pretend it never crashes, in practice every time I run a kernel I've compiled it seems shaky, and ubuntu doesn't always behave. I should perhaps instead claim that the tested-to-death debian stable never crashes. Anything less rigorously tested is unfortunately a bit less trustworthy
Click to expand...

Which is why we employ lots of System Analysts on exhorbitant fees.:D (and charge accordingly of course.) Have fun with the Z-series google.

What are you studying by the way, I see engineering, what discipline?
 
Ok, i did 6 hours of Prime, on stock settings fine with no errors. I have now loaded back up my 4GHz OC profile and set the Ucore to x16, and at the moment Prime is coming up to the 2 hour mark. My current voltage settings are:

Vcore 1.216-1.232 (CPU-Z)
QPI as 1.275 (BIOS)
IOH Core 1.14 (BIOS)
CPU PLL 1.3 (BIOS)
DRAM 1.640 (BIOS)


Currently at these voltage leves Real Temp is reading temps of, 74,74,73,71. Do these seem a bit on the high side considering i have a Corsair H50 with push/pull setup??

Regarding the CPU PLL, a lot of people have said to lower this from the default setting, would this make the system less stable though??,
 
Your temps are fine. About what to expect with your vcore and QPI voltages. Reducing the CPU PLL does seem to improve stability strangely enough.

I didn't realise you had your uncore set to x17, on the UD5 this would cause Prime to fail. x16 is what it should be for 1600mhz memory.
 
Gaidin109 said:
Your temps are fine. About what to expect with your vcore and QPI voltages. Reducing the CPU PLL does seem to improve stability strangely enough.

I didn't realise you had your uncore set to x17, on the UD5 this would cause Prime to fail. x16 is what it should be for 1600mhz memory.
Click to expand...

Is that right?, I thought you wanted the uncore at x2 the speed of the ram, so as near to 3200 as possible which was what x17 game me over x16, but the system does seem to be priming better now, and if that's what you recommend.

Now I'm on x16 uncore would I be able to get my vcore and qpi lower than what I had before, as these were the lowest I could get them when on x17 uncore, any lower and LinX would fail
 
Last edited:
lettuce said:
Is that right?, I thought you wanted the uncore at x2 the speed of the ram, so as near to 3200 as possible which was what x17 game me over x16, but the system does seem to be priming better now, and if that's what you recommend.

Now I'm on x16 uncore would I be able to get my vcore and qpi lower than what I had before, as these were the lowest I could get them when on x17 uncore, any lower and LinX would fail
Click to expand...

x16 gives you 3200. You might be able to get a lower vcore and qpi if the uncore was the only thing causing the failure. Try it and see. One at a time tho or you won't be able to tell which is failing.
 
Ah ok then, so I'm basically been overclocking my memory then at x17, and obviously by too much hence prime failing
 
lettuce said:
Ah ok then, so I'm basically been overclocking my memory then at x17, and obviously by too much hence prime failing
Click to expand...

basically yes, although its the qpi link thats being overclocked not the memory, but because its outside of the spec its causing an instablility in the IMC thus the crash.
 
JonJ678 said:
That's quite interesting Gaidin. Any chance you have a spare hard drive/usb stick you could put a virgin copy of windows on to see if you can replicate the error? I'm assuming your copy of prime has the same md5 checksum as listed.

I tend to reinstall windows repeatedly while overclocking to reduce the chance of software being the issue, and generally the overclocking install just has windows and the tools on. I can't help wondering if 'no system is ever 100% stable' is a consequence of using windows too much :p

I'd suggest 9-9-9-28-2, and also try 8-8-8-24-2 at the x6 multiplier. I think the issue you have is the imc giving up on you when pushed too hard, which is most likely solved by feeding it more qpi or asking less of it. Gaidin makes a good point though, it may well be software related. Could you try with a fresh windows install (installed at stock speeds, of course)?
Click to expand...

I used an updated version of Prime, going from v25.9 to v25.11 and passed 8 hours of blend no problems whatsoever, ran it twice with the same result. Kinda proves it was the Prime software rather than the OC.
 
Well it appears i cant get any lower than:

Vcore 1.268
QPI 1.275
IOH Core 1.14
CPU PLL 1.3
DRAM 1.640

I would have thought i could have got the QPI voltage down more than that. Any other tweaks i could do to get lower Vcore & QPI voltages??
 
It suggests the version of prime was corrupted to me, or a software incompatibility which I know too little to diagnose. I agree its looking like software.

Those look good lettuce, I feel you should be aiming for higher speeds rather than lower voltages at this point.

Mechanical engineering :)
Thermofluids really should be applicable to water cooling, but I don't seem to be having much luck there
 
Thermofluids is a branch of thermodynamics that deals with the heat flow and energy transfer of liquids, gases and vapour is it not?
Most refrigeration techniques use this principle, vapour compression or dry ice. Vapour absorption uses water and ammonia to some effect. Most water-cooling systems can be said to use these principles.

Now computational fluid dynamics is really cool, simulating wind tunnels and water flow models is cool stuff. IMO


Mechanical Engineering, cool stuff. What level are you studying at....
 
Last edited:
It's about time there was a VCC system for PC's, like the modular refrigeration units and rear doorheat exchangers we use on mainframes to sustain sub-ambient temps in multi-chip-modules.
 
Undergraduate so far :)

vcc = variable climate control, or something else? A fair few people run phase 24/7, and I personally want to build a tec chilled refrigeration chamber containing a desktop. The big issue there is funding, as a student can't really drop several hundred on a fridge.

I'm curious about the two applications mentioned, how cold do they run the case ambient relative to room temperature?
 
VCC= Vapour Compression Cooling, not unlike Phase Cooling but with much slower mass flow rates. The cooling is directly linked to the amount of power introduced to the refrigerator, usually the coefficent of performance is around 2-3. Set temperatures depend on application, amount of rack space, cpu utilization etc... SAN's are 0c-40c with relative humidity of 20%-85%.
In the z-series they have a closed coolant loop that has redundant distribution units at the bottom of each rack.

When you think that the average rack is putting out 75000w of heat then you can see its pretty hot, with the Rear-Door heat exchangers (basically it works like a car radiator) you are looking at the exchanger taking out 110% of the heat generated by the rack, this means the air exhausted is actually cooler than the air going in.

For something really cool:

The Hydro-Cluster Power 575 series supercomputers can actually be configured to allow data-centres to capture the excess heat and export it as hot water for energy use off site and even sell it on. Crazy stuff.

Using a tec-chilled refrigeration unit for your PC is not easy, if the dew point is below that of the surface temps then condensation will kill everything. You would have to insulate all the surfaces from air. this is why we only cool the processor modules with refrigerated units and the rest of the rack is aircooled.


Last one installed had a 18c controlled ambient room temperature and 5.5c rack temp. Its dependant on lots of variables including pressure and altitude etc. but will operate within a 10c to 32c ambient range and 8% to 80% humidity.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom