Haswell overclocking discussion.

[guskline]

Thanks Gotta luv a good build in winter!

Usually allow 48hours to stress test a decent project build, typical software used are Prime95 , Intelburntest , Unigine , 3Dmark , FurMark, CrystalDiskmark.
Monitoring software- Speedfan, GPUz, CPUz, Taskmanager.
Diagnostics software- Whocrashed , Eventviewer..

Thanks for the info!

 

[xtruder]

My 1st time using Cinebench, 1st attempt is not any better than yours. And my CPU dropped from 4.6ghz to as low as 3.7ghz during the CPU test? and the Vcore increased from 2.8v to 2.96 which suggest the CPU needed the extra juice under load but the temps stayed the same, quite odd.. Took a screenie of it while in action..


Anyone know why the drop in CPU frequency while running the CPU test?

You are probably throttling due to high temps, what are they?

Throttling in cinebench is pretty bad, usually it shouldnt exceed 60-65C, unless you are on stock cooler running at insanely high volts.

Btw, the results you are quoting are from 4670k. With 4770k, if it stayed at 4.6GHz, you should be getting a score of 10.15+.

 

[John Tauwhare]

This is an old thread but it seems the most appropriate place to post my test results. I'm just very late to the party.



Much of the testing is not optimal as there are so many extra permutations with the Uncore clock and voltage, and the Vccin, and I've not run P95 for hours to confirm stability, but I'm getting a good feel for the chip.

Stability is all relative anyway, but by producing a profile of results under different tests I've tried to illustrate the whole envelope. I'm testing the chip here, not the box, which is compromised by it's small mobo and generally rubbish airflow. So, I've left the lid off and had it pointed at an open window, hence the low ambients.



I'm chuffed I got a good chip. Not as good as Balla's, but it's better than average. And it has taken 1.4 seconds off Face2Face's Excel benchmark record :biggrin:.

 

[Face2Face]

^ Welcome back John. I see you just came back to stomp my Ivy... Nice little setup btw, are you using it as a HTPC? I have been itching to build one myself.

 

[ShintaiDK]

Nice box John. Viva La MiniITX!

 

[John Tauwhare]

^ Welcome back John. I see you just came back to stomp my Ivy... Nice little setup btw, are you using it as a HTPC? I have been itching to build one myself.

Thanks. Mission accomplished, just before year end . I will use it as an HTPC when I'm finished playing with it. But look what Santa brought me..

 

[John Tauwhare]

Nice box John. Viva La MiniITX!

Thanks Shintai. Thinking maybe I should have got the SG08 box, but it's not too late.

 

[BallaTheFeared]

Thanks. Mission accomplished, just before year end . I will use it as an HTPC when I'm finished playing with it. But look what Santa brought me..

lamo!

Thankfully I never needed one myself, probably would have destroyed it like another user I know :whiste:

 

[Morbus]

I have actually destroyed a couple of those with a hammer. If you don't know what they're for, my advice is do not touch them.

 

[jj109]

I have actually destroyed a couple of those with a hammer. If you don't know what they're for, my advice is do not touch them.

'Those' as in the big blue thing in the box? Oh no, that's not what he's talking about.

 

[John Tauwhare]

I just did the delid on my 4670K. I tried with a blade but couldn't get one in, so I used a vice, and that was easy. Made a video..

http://www.youtube.com/watch?v=3CX8rsnztro
(still trying to get the youtube link to work)

Under the IHS was a surprise. Extremely thin layer of TIM, even thinner that it looks in the photo, just a smear. And the die surface has only residue leaving it shiny.



Must partly explain why benching temps were good and suggests there is little left to be gained. What they hey, I had to know..

The PC case is in the shed being painted. When it's dry I'll reassemble everything (using CLP as TIM) and report back on temp changes - assuming the shock of hammering didn't kill the CPU.

 

[BallaTheFeared]

Your video link is censored, is it youtube?


I don't think you can break it with shock, you need to damage it physically to kill it.

 

[John Tauwhare]

Your video link is censored, is it youtube?

I don't think you can break it with shock, you need to damage it physically to kill it.

Link (above) seems to be working now. My first experience of making a youtube video.

 

[BallaTheFeared]

You have pretty hands, a/s/l?

lol

Looks painless, can't believe someone who shall go unnamed managed to mess it tviceman.

 

[John Tauwhare]

You have pretty hands, a/s/l?
lol

Lmao! Not as pretty as my little blue PC?

 

[John Tauwhare]

POST 153
This is an old thread but it seems the most appropriate place to post my test results. I'm just very late to the party.



Much of the testing is not optimal as there are so many extra permutations with the Uncore clock and voltage, and the Vccin, and I've not run P95 for hours to confirm stability, but I'm getting a good feel for the chip.

Stability is all relative anyway, but by producing a profile of results under different tests I've tried to illustrate the whole envelope. I'm testing the chip here, not the box, which is compromised by it's small mobo and generally rubbish airflow. So, I've left the lid off and had it pointed at an open window, hence the low ambients.

After the delid I cleaned up the IHS and PCB, mainly with the edge of a credit card and fingernails. Then I put it back together with CLP on both IHS and PCB, and a very thin layer of silicon sealant around the outside of the IHS. Put it back in the motherboard and reattached the waterblock with CLP on both surfaces.

Massive relief when I switched on and the 4670K still worked.:thumbsup:

Then I ran all the tests again, with the same clock and voltage settings, but higher ambients:



The temp reductions were much more than I'd expected. As much as -34C!!

Allowing for the increase in ambients the changes across the 4 most stressy tests were -36C, -35C, -32C, -32C. And all three of the Cinebench, Excelv1 and Excelv2 benchmarks had improved scores (on the first run of each).

So delidding of Haswell, and application of CLP - even of an IHS which has been quite well put on by Intel with a very thin layer of original TIM - is still worthwhile.

 

[pm]

The temp reductions were much more than I'd expected. As much as -34C!!

Allowing for the increase in ambients the changes across the 4 most stressy tests were -36C, -35C, -32C, -32C. And all three of the Cinebench, Excelv1 and Excelv2 benchmarks had improved scores (on the first run of each).

So delidding of Haswell, and application of CLP - even of an IHS which has been quite well put on by Intel with a very thin layer of original TIM - is still worthwhile.

These results are consistent with my delidding experience as well - although your testing methodology is much more thorough than mine. But I saw a ~30C reduction in peak temps as well.

 

[JoeRambo]

Nice temps. Maybe you can try latest Linx (with latest Linpack) on 4.6Ghz/1.3v setting? 118Gfps can barely warm a powerhouse like Haswell

 

[BallaTheFeared]

Forget Linx, run Intel Optimized Linpack.

Can you break 200 GFLOPs?

 

[JoeRambo]

Forget Linx, run Intel Optimized Linpack.

It's actually same thing, as Linx is just a wrapper for Linpack. Much more convenient than to mess with command like, manual start command CPU affinities when on HT etc

 

[BallaTheFeared]

Pretty sure last time I used linx it gave me 140 GFLOPS vs 220 with Intel's.

Has that changed?

 

[JoeRambo]

Pretty sure last time I used linx it gave me 140 GFLOPS vs 220 with Intel's.

Has that changed?

It got updated to latest Linpack executables. Linx is just an convenient UI that is running Linpack binaries. And you can paste latest Linpack binaries into linx 0.6.5 yourself and it will just work.

 

[John Tauwhare]

Forget Linx, run Intel Optimized Linpack.

Can you break 200 GFLOPs?

I'm not familiar with LinX or Linpack and I've only had a few minutes to do this but I've downloaded LinX. Initial run at 40/38 gave 178Gflops. Quickly discovered that the stress is increased by the problem size and memory. Maxxed out the setting to 6Gb memory and started it. Stopped when power draw reached 155W! (but still only 80C). Ran it 5 times at 46/38 for 20k problem size and 3Gb memory and got this:



Temps 73/79/78/77 and maxxed at 140W. With a more time I could get more out of it. Or kill my chip? Good result though and I'll leave it at that for today.

 

[JoeRambo]

Temps 73/79/78/77 and maxxed at 140W. With a more time I could get more out of it. Or kill my chip? Good result though and I'll leave it at that for today.

Those are GREAT temps at that voltage, Linpack with Haswell FMA support is massive load, temps will not break 50C in gaming

 

[BallaTheFeared]

I stand corrected on LinX.

Very nice, I have to run my Corsair SP 120s at 100% to do it.


Bump your ram up for a few more flippy Floppies