DFI UT X58-T3EH8 continued - Part 2

[Infa]

Nice cpu! Nice work I would push it more though say 4,6?

Like the the way you have put this together clean and tidy and just the info that needs.

regards.

[pangingIII]

Thanks Infa.

[jcool]

jcool praz is right

this is a known fact and was discussed even during the x38/x48's days
that the cpl chip doesnt need to be cooled and theres no internal temp reading from them..

almost everybody who blew their board after replacing the tim on their pwms
was because they had too much contact with the cpl chip.

Sure, if you say so - I only do a customer build with the DFI every other week, what do I know...

My observations leave zero room for interpretation - the cpl chips overheat if they are not properly cooled. You guys just don't notice because most here probably have 2000RPM case fans, CPU fans etc. With case/rad fans running at 500-600rpm and water cooled CPU, it just does not work w/o cooling the big fat suckers

[cstkl1]

Sure, if you say so - I only do a customer build with the DFI every other week, what do I know...

My observations leave zero room for interpretation - the cpl chips overheat if they are not properly cooled. You guys just don't notice because most here probably have 2000RPM case fans, CPU fans etc. With case/rad fans running at 500-600rpm and water cooled CPU, it just does not work w/o cooling the big fat suckers

nope nothing on it

heck even worse actually cause i have a coolit freezone at the top of the cpu...
no fan at all.

my observation is based on having 4 dfi'x x58's two of them running in a watercooled case rig and two for bench. no need at all.

to reduce the pwm temps i have a hr05 +70mm fan on it.
thats about it.

[OnDborder]

PangingIII, what ram are you running?

[pangingIII]

PangingIII, what ram are you running?

Crucial value 2x2GB ram kit.



i use 3 kits, total 6x2GB 12GB

[Praz]

You have any idea what could die so easily on this board?

Hard to say. You don't know what was replaced last time?

[Peen]

Hard to say. You don't know what was replaced last time?

The DFI tech said a chip was burnt in the PWM area

[pokipoki]

Does this board support Xeon E5520? I mean does the bios allow unlocked multis, OC & memory divider options with this xeon?

The main reason I prefer E5520 over W3520 is the heat output, 80W vs 130W.

[ChaosMinionX]

Has anyone had better luck using different BIOS's for their chips? I have yet to have any such luck with a C0/C1

[cstkl1]

Has anyone had better luck using different BIOS's for their chips? I have yet to have any such luck with a C0/C1

most of the bioses are more dram chipset related..

nothing to do with proc overclocking ..

[donmarkoni]

Has anyone had better luck using different BIOS's for their chips? I have yet to have any such luck with a C0/C1

What BIOS are you using?

most of the bioses are more dram chipset related..

nothing to do with proc overclocking ..

I had different experience. Using 2/17 and 2/27 CPU needed 1.36V VTT set (1.376 read) for 4GHz (200x20). With 5/08, 6/19 and now 7/24 it needs only 1.29V VTT (1.312) for 4-4.2GHz (200x20-21) and 1.31V VTT (1.328) for 4.2-4.4GHz (210x20-21)

[cstkl1]

What BIOS are you using?



I had different experience. Using 2/17 and 2/27 CPU needed 1.36V VTT set (1.376 read) for 4GHz (200x20). With 5/08, 6/19 and now 7/24 it needs only 1.29V VTT (1.312) for 4-4.2GHz (200x20-21) and 1.31V VTT (1.328) for 4.2-4.4GHz (210x20-21)

again i repeat thats just because of the RTL and skew strengths for mem chipset for the uncore..
again mem related.
proc speeds are nodiff.
overall the new bios takes a huge hit on latency and mem bandwidth.

for all the bios for under 1700mhz mem speed my vtt is 1.21v right up to 4.5ghz for elpida hypers
so again it depends on ure mem

[donmarkoni]

again i repeat thats just because of the RTL and skew strengths for mem chipset for the uncore..
again mem related.
proc speeds are nodiff.
overall the new bios takes a huge hit on latency and mem bandwidth.

for all the bios for under 1700mhz mem speed my vtt is 1.21v right up to 4.5ghz for elpida hypers
so again it depends on ure mem

You are right, but it definitely helped, lowered temps for sure a bit, and now I'm in Intel's specs safe range.

[ChaosMinionX]

Im using 7/24.... I have been trying to get this chip down, but it seems to be a dud for 4ghz range, at least as far as prime95 is concerned.

[sjakie666]

Hi,

Is there a way to keep the turbomode with thermal management enabled???

thnx

[donmarkoni]

Im using 7/24.... I have been trying to get this chip down, but it seems to be a dud for 4ghz range, at least as far as prime95 is concerned.

I wasn't paying attention, did you post your bios settings anywhere? I would like to compare it to my, cause I have C0 also.

Hi,

Is there a way to keep the turbomode with thermal management enabled???

thnx

Nope. AFAIK

[sjakie666]

dammit..... a well, than away with Thermal M.....

[donmarkoni]

dammit..... a well, than away with Thermal M.....

Just use Shutdown temperature or something similar. I disabled thermal management, but set shutdown temp @ 100 deg.C, to be safe.

[jcool]

Hi,

Is there a way to keep the turbomode with thermal management enabled???

thnx

Sure. TM enabled only throttles when a certain temp is reached, I haven't checked on the UT but in my DFI x58 JR that temp is around 85C.

But like donmarkoni said, you can disable TM and use your own shutdown temperature to be sure it won't throttle.

[sjakie666]

wierd.... because my 21x mult. is jumping with TM enabled. my cores don't get higher than 70 degrees......100% stressed. (saw it with i7turbo)
Does thermal management also keep a eye on pwm temps and IMC temps?

[cstkl1]

memtest86+ 3.0

approved..
finally a good memtest to find mem skew aka bootup bclk
vtt voltage
and dram voltage..

what u get here is the same as prime.

[p2501]

Say guys is there something else than prime or linx to load up memory really heavy to test for mem/imc problems? Seems I might have a problem that none of both can catch.

[zalbard]

Say guys is there something else than prime or linx to load up memory really heavy to test for mem/imc problems? Seems I might have a problem that none of both can catch.

Why not use Memtest?

[Sam_oslo]

Say guys is there something else than prime or linx to load up memory really heavy to test for mem/imc problems? Seems I might have a problem that none of both can catch.

Have you tried HyperPi32M? It is very "picky" and accurate test for MEM/IMC-stability on my setup. But it is not as accurate on Core stability, tho.