Does XTU support AVX / AVX2? If not, it's not really useful as a stress test for Haswell. The only way to truly stress test Haswell is with AVX instructions. Everything else will be 30C lower and will not require the same voltage.

 

I don't think 90C is too high in the pure AVX FPU test. As a reference, a 4770K at stock with the stock cooler will hit about 86C in the same test, so these kind of temps are within reason.

When the temp hit 100 and throttling started at 1.3 vcore and 4.6Ghz, that was on Prime 95 blend, so I would have hated to see what the FPU only test would have been.

 

Mind you I was just tinkering with settings so there was definitely room to make some definite improvements.

 

XTU does support AVX and seems to be fairly similar to AIDA in how it tests.

Xtu 4.1 has the benchmark , hard to find what instruktions the benchmark use ,

Probaly they use avx this bench is for Z77 Z87 boards only , try it and report back our toughts .

 

Yes a run 2v minimum for uncore mem OC 1.18-1.22v

Hmmm haven't tried setting VRM quite that high. Maybe I'll give that a go and push for 4.6 again My only real concern with pushing the voltage is the temps.

 

With CPU multi at 45x, my uncore is 42x at 1.15 uncore voltage and temps are all in check. It seems like a big voltage jump to get to 4.6Ghz.

I've got my 4670K sitting at 4.4GHz @ 1.155 VCore.  Temps idle in the low 30s.  I've only got a cheapo Coolermaster Hyper 212 EVO for a cooler, so I don't want to push my CPU further.  The next thing I need to do is get into overclocking my RAM and GPU.

 

For stability tests I ran IBT for about 3 hours and Prime95 for 4 hours with no errors to find my stability.  IBT was obviously the hardest on temps and my hottest core hit 88C - the rest were in the low 80s.  The max temps I have hit during gaming, including FSX with heavy add-ons, has been 61C.

 

I've been using the PC on these CPU settings for about 2 months now and haven't had any issues.  I don't need to run stability tests for endless hours for my personal satisfaction, as long as it does what I want it to do without any issues than it is fine for me.  I use it for work, photoshop, and gaming.  I've thrown every game I have at it on Ultra settings and it eats them alive with no problems.

XTU wasn't very stressful on my system. Temperature in the 60's range at 4.5 GHz.

 

 

Remember there's also the brand new Linpack AVX2 test:

http://software.intel.com/en-us/articles/intel-linpack-benchmark-download-license-agreement

 

This requires you to turn off HT for it to run properly, then run the Xeon64 .bat file. Your temps will be about 10C higher than Aida64 FPU-only. When this was posted at another hardware forum, almost everyone who ran it had to take their OC down to 4.2 - 4.3 GHz to run it without throttling, even those who were at 4.8 GHz normally. It's always possible to artificially stress the CPU to the breaking point. Have to draw the line somewhere.

Forget the stress tests. Just use your puter normally with the highest stable clock u can get whilst the temps staying in control.

XTU wasn't very stressful on my system. Temperature in the 60's range at 4.5 GHz.

 

 

Remember there's also the brand new Linpack AVX2 test:

http://software.intel.com/en-us/articles/intel-linpack-benchmark-download-license-agreement

 

This requires you to turn off HT for it to run properly, then run the Xeon64 .bat file. Your temps will be about 10C higher than Aida64 FPU-only. When this was posted at another hardware forum, almost everyone who ran it had to take their OC down to 4.2 - 4.3 GHz to run it without throttling, even those who were at 4.8 GHz normally. It's always possible to artificially stress the CPU to the breaking point. Have to draw the line somewhere.

Do you run the xtu stresstest or xtu benchmark? The stresstest is not as hard as the benchmark you have the benchmark in xtu 4.1 and later

I think my uncore 42x at 1.15v wasn't actually stable as I had a crash in Prime after 4 hours. I'm upping the voltage and testing again.

 

It's all a bit trial and error but I think that could have been the cause of my instability at 4.6Ghz as well.

 

Here we go again! lol

Hi

Run some AIDA FPU stress test after delidd had only MX2 but that worked to.

hope to do 4.9 when a have CL Pro.

 

Westman that's great results for the FPU only test at that speed/vcore. Can't wait to see what you can fo once the CLP arrives.

 

I'm calling stable at 4.5Ghz for the time being but will continue to try push higher. I think temps will probably be my only limiting factor since I don't plan to delid.

 

Here's my settings etc.

 

CPU Model: 4770k
Core Multiplier:45
Vcore: 1.225
Uncore Multiplier: 42
Uncore Voltage: 1.18
VCCin:1.82
Stability Test: Prime 95 (24 Hours), IBT, AIDA64, Cinebench, Handbrake, General Use
Batch Number: L307B246
Ram Speed: 2400Mhz XMP

L307......That chip will run 5.0GHz @ 1.44V.

 

HLJAMES

Really? I guess I got one from a good batch then?

 

I guess if I want to go that far I'll have to take a look at delidding and probably a custom loop.

 

Unfortunately that's not going to happen any time soon though.

Just to add the to OP topic:

 

CPU Model: 4770k
Core Multiplier: 45
Vcore: 1.19v
Uncore Multiplier: 43
Uncore Voltage: 1.15v
VCCin: Auto (1.78v)
Stability Test: OCCT 4.4.0 CPU (2 hours), FSX (several hours, not a single BSOD, lockup or CTD), LinPack/Prime95 AVX ~30 minutes each, on multiple occasions
Batch Number: L315B371
Ram Speed: 2133Mhz

Notes: HT disabled, see "My PC" for more details

 

Obvious I've not tried to beat my OC to death with stability tests, though I've had plenty of BSODs and lockups getting to where I'm at; feel free to brand me as a charlatan for being so unconcerned.  My bottom line is how stable it is in FSX and so far its been 100% rock solid. I've toyed around a bit running HT which required 1.21v using the above settings, though I yet to give it any decent amount of stability testing beyond 30 min of OCCT CPU and a little LinPack/Prime95 AVX.  I'm not motivated to dial in HT since my OC is 100% dedicated to running FSX; HT really bumps up the stability testing temps so IMO it's not worth the potential damage for something of which I have no use (at least at this time).  I too at this time plan not to delid.

Disabling HT is fine if you don't need it. After all, the 4670K doesn't have HT, and we don't consider all 4670K owners cheaters   However if you can do 4.5 GHz at just 1.19V, I don't think you'd have any temperature problems with HT enabled, either. Generally I've found there's up to a 10C penalty for enabling HT.

 

With such a CPU I would be tempted to delid and see what it can do at 1.3 - 1.35v. Might be able to break 4.8 GHz or even 5 GHz...However Haswell @ 4.5 GHz is nothing to sneeze at in terms of performance, and with that low core voltage, power usage will be kept down as well.