Overclocking on a i7-4790S?

[LTC8K6]

3.8GHz should see over 200GFLOPS. Some locked Haswell's can be unlocked, others can use the top turbo bin on all cores when using microcode version 0x7 or less.

200Gflops in what benchmark?

 

[LTC8K6]

With IBT, my 4790K was around 120GF at 4.4Ghz, the 1231 V3 was comparable at just over 100GF at 3.8Ghz.

 

[Dufus]

Linpack for which IBT is a wrapper. Perhaps the version of Linpack used is an old one.

Best performance when number of threads = number of cores and each thread is run on a different core.

Example running less than 3.8GHz

Of course being a laptop cooling is a big problem with consequent power consumption and heat generated.

 

[LTC8K6]

Getting 145 with 8 threads and 176 with 4 threads.

No throttling that I can detect so far. Your Bus speed is a little higher, and your ram is a little better.

 

[LTC8K6]

Bumping the bus to 102/ram to 950, didn't increase the numbers much.

No way I'm getting anywhere near 200GF at 3.8Ghz with this Xeon.

I do have higher residual numbers, whatever that means.

 

[Dufus]

For 15000 problem size and 4 threads my residuals are always the same. Try this one You may need to manually set affinity on Linx before benching. Also with those residuals try running 3 runs. Any monitoring software may affect performance. Check clocks while running too.

 

[LTC8K6]

Well, there's virtually no difference between our processors and ram, you have 2mb less cache, that's about it.

Their performance is going to be pretty much identical at the same clock speed, so it has to be other factors such as Windows version or background programs or something else.

LinX versions are the same, apparently, despite the different look of our screen caps.

 

[Zodiark1593]

My i5-4590 scores 160 Gflops in LinX running at 3.7 GHz. I run 1600MHz RAM though so that may impact my numbers vs the Xeon more so than the 100 MHz difference (also had task manager and cpu-z open). My BCLK also seems a tad low too at 99.8 MHz. Wonder if I could up that to 101 MHz on an H97 board.

 

[LTC8K6]

My i5-4590 scores 160 Gflops in LinX running at 3.7 GHz. I run 1600MHz RAM though so that may impact my numbers vs the Xeon more so than the 100 MHz difference (also had task manager and cpu-z open). My BCLK also seems a tad low too at 99.8 MHz. Wonder if I could up that to 101 MHz on an H97 board.

That's pretty well in line with my Xeon score of 176GF at 3.8Ghz.

Don't know how the 4700MQ is doing so much better than us.

 

[Zodiark1593]

That's pretty well in line with my Xeon score of 176GF at 3.8Ghz.

Don't know how the 4700MQ is doing so much better than us.

Hyperthreading can deliver about 33% additional performance in certain, heavily threaded tasks. Haswell seems particularly good at utilizing it. I'm actually surprised it's not higher since it should maintain 3.7 GHz for 20 seconds, though long term workloads, the desktop i5s will pull ahead.

 

[LTC8K6]

Hyperthreading can deliver about 33% additional performance in certain, heavily threaded tasks. Haswell seems particularly good at utilizing it. I'm actually surprised it's not higher since it should maintain 3.7 GHz for 20 seconds, though long term workloads, the desktop i5s will pull ahead.

Well, we are only using 4 threads, and my Xeon and his 4700MQ are both Haswell 4C/8T chips running at approximately the same speeds.

My Xeon E3 1231 V3 is essentially an i7-4770 without the graphics.

I am maintaining 3.8Ghz forever, and I'm nowhere near his score of 200.

So there's something different between the benchmarks.

I should be defeating the 4700MQ in the benchmark, yet I am not even in the same ballpark.

 

[LTC8K6]

2133 ram:

 

[Zodiark1593]

With my RAM speed upped a bit and off a fresh boot, I scored 23 seconds (and 12.969 in Time) and 173.5 Gflops.

I would be posting pics, but that isn't working for some reason.

 

[LTC8K6]

Closing in on it:

 

[Dufus]

@LTC8k you do mean Linpack version was the same and not just Linx version, right?

Only strangeness with my CPU is that with AVX loads it does not overvolt unlike some.

 

[LTC8K6]

@LTC8k you do mean Linpack version was the same and not just Linx version, right?

Only strangeness with my CPU is that with AVX loads it does not overvolt unlike some.

Not sure, I never did download the one you suggested.

 

[LTC8K6]

When I put my 4790K back in the Z97-A, it gave me the "dead CPU" LED on the board, and wouldn't do anything.

Moved the 4790K to another board. No go.

3 other Haswell chips work fine in both boards.

So apparently, my 4790K is now deceased, after less than one year of occasional use.

Update: RMAing it to Intel today.

 

[DrMrLordX]

That sucks. It just coughed and died? Don't let it bite you when it crawls out of the package and attacks you.

 

[Dufus]

+1
Strange how it died. Kind of reminds me of some of the ASUS P67 board bugs one which when changing profiles would set CPU offset voltage to greater than 0.8V, meaning the possibility of over 2V to the CPU.

 

[iSkylaker]

I guess this explain why they can be overclocked
http://www.techpowerup.com/202196/a...-h97-h87-b85-and-h81-series-motherboards.html

 

[LTC8K6]

I don't know what happened to the 4790K. I was not even using it for these tests. It was fine the last time I used it, and the next time, nothing.

UPDATE EDIT: Got a tracking number for my replacement 4790K. Woo hoo!

 

[LTC8K6]

Just as a follow-up, the replacement 4790K works perfectly.

 

[LTC8K6]

hey lol i wish i knew about this post earlier. any questions about the video or tests you want me to run or settings?

i guess i would like to know any other things i could do to my system to get the most out of it. i've been working on this series of videos also about optimizing your computer for responsiveness in gaming and just general stuff https://www.youtube.com/watch?v=wZ31fKQUWnQ

You are slightly behind my Xeon 1231-V3 running at 3.8ghz, so you have work to do.

 

[LTC8K6]

i had stuff running like chrome and my memory is mismatched so i can't go higher with my memory speeds >..< 100:100 ddr3-1400 is what i was at when i ran that. hyperthreading disabled for reasons atm. but earlier in this thread people said stuff about throttling. zero, none to be seen anyplace. thermal or TDP. i disable power saving crap on everything i got now. my r9 390x and my cpu both run at full clock most of the time. well my cpu does but at night i lower the 390x down.

Yeah, I was running CL10 2133 ram with the Xeon.

It's essentially a 4770K without the graphics for about $50 less.

 

[Magic Carpet]

I don't know what happened to the 4790K. I was not even using it for these tests. It was fine the last time I used it, and the next time, nothing.

A post-mortem examination will show. I wonder if it's going back to Intel.