2600k and p8p67 deluxe - any bets? any advice?

[daddylongs]

Care to take and bets on how high I will get (stable) with the below parts?
I am curious because this is the most research and $$ I have thrown towards a system to date.

Asus P8P67 Deluxe
SB 2600k
Mugen 3 (SCMG-3000)
DDR3-2000 (Gskill Trident 2x2GB)
Antec 300, with 2x extra Scythe 120mm 900rpm front fans.

I would have been up and running by now but the first mobo newegg sent me had obviously been bent by a prior user. At least the cpu was new.

Care to take any bets on how high I will get?

Any concerns or tips? I had great luck with this case and RAM with my c2d and ep45c-UD3R.

 

[DominionSeraph]

9GHz at an altitude giving 3ATM on Neptune.

Air cooling FTW.

 

[daddylongs]

9GHz at an altitude giving 3ATM on Neptune.

Air cooling FTW.

:whiste:

Given that I dont want to push the voltage above 1.4 for the first few weeks, I was just curious really what my odds are of reaching 5Ghz stable....

 

[Diogenes2]

It will take at least 1.45 - 1.5 to get 5g, unless you have a very rare chip ...

 

[MrTransistorm]

No bets and no guaranties. Every CPU is different.

The chances of getting 5 GHz with Vcore less than 1.4V are very slim. Most need at least 1.45V.

I don't know much about that Mugen 3, but don't expect to get 5 GHz on air either. Your temps will probably be too high.

What PSU are you using? It seems that in almost every post I see like this one, the PSU is rarely mentioned. This implies that it is an afterthought. It bears repeating that the quality of the PSU is very important. A cheap PSU may cause instability under load and when booting up. If it fails, it has the potential to damage every other part in the system. If you stick with brands like Seasonic, Corsair, and Antec, then you are much less likely to encounter problems down the road.

 

[MadScientist]

1.65V ram. You can use it but I'd suggest lowering the voltage.

 

[Idontcare]

Care to take and bets on how high I will get (stable) with the below parts?

Define what "stable" means to you.

To me "stable" means 5 passes of LinX on all-mem for a minimum, 20 passes for verification.

5GHz stable is a challenge unless you don't mind near-90C temps and near-1.5V Vcc's.

 

[daddylongs]

No bets and no guaranties. Every CPU is different.

The chances of getting 5 GHz with Vcore less than 1.4V are very slim. Most need at least 1.45V.

I don't know much about that Mugen 3, but don't expect to get 5 GHz on air either. Your temps will probably be too high.

What PSU are you using? It seems that in almost every post I see like this one, the PSU is rarely mentioned. This implies that it is an afterthought. It bears repeating that the quality of the PSU is very important. A cheap PSU may cause instability under load and when booting up. If it fails, it has the potential to damage every other part in the system. If you stick with brands like Seasonic, Corsair, and Antec, then you are much less likely to encounter problems down the road.

This is a great point, but I didnt want to scare people away from my post with too much info. I have learned my lesson the hard way in the past about PSU quality.

This time around I have an Antec EarthWatts 750w. It was a random gift so I am hoping it will work for now. I am happy to upgrade if it is going to endanger my setup.

Define what "stable" means to you.

To me "stable" means 5 passes of LinX on all-mem for a minimum, 20 passes for verification.

5GHz stable is a challenge unless you don't mind near-90C temps and near-1.5V Vcc's.

Stable means perma-stable under load - like I can come home and play some games for a 1-4 hours without having to worry about issues, or without having to tinker with settings. Being able to do some folding on this side would be nice too.


love you guys, keep it coming!

:biggrin:

 

[daddylongs]

Define what "stable" means to you.

To me "stable" means 5 passes of LinX on all-mem for a minimum, 20 passes for verification.

5GHz stable is a challenge unless you don't mind near-90C temps and near-1.5V Vcc's.

Sorry for the double post - can you give me a rough idea of how long those 20 passes might take? 3 hours? 3 days? I've been out of the OC ring for a couple years and my standards are old history.

 

[Idontcare]

Mugen 3 (SCMG-3000)

Sorry for the double post - can you give me a rough idea of how long those 20 passes might take? 3 hours? 3 days? I've been out of the OC ring for a couple years and my standards are old history.

http://www.rwlabs.com/article.php?cat=&id=454&pagenumber=6

That cooler doesn't seem to be all that highly rated compared to the rest of the field, this will limit your OC's for sure.

At 4.5GHz on a 2600K one LinX pass takes around 8-9 minutes. But I don't know if you'll get there with your HSF, it won't hurt to give it a try though.

Worst case is that your temps go to 90C and you'll see that right away, in which case you would then back down on the clocks.

 

[daddylongs]

http://www.rwlabs.com/article.php?cat=&id=454&pagenumber=6

That cooler doesn't seem to be all that highly rated compared to the rest of the field, this will limit your OC's for sure.

At 4.5GHz on a 2600K one LinX pass takes around 8-9 minutes. But I don't know if you'll get there with your HSF, it won't hurt to give it a try though.

Worst case is that your temps go to 90C and you'll see that right away, in which case you would then back down on the clocks.

You're right, it really is just a mediocre cooler when it comes to performance. Xbitlabs often uses the Mugen2 for their mainstream OC reviews, so it seemed like a safe bet.

Do you (or anyone else) have any suggestions on a cooler to use that will fit my case (antec 300) and mobo (p8p67 deluuxe), and not break the bank, and be pretty quiet? I was thinking about the MUX-120 but I was afraid it would be too tall. I'll take some measurements once I get the Mugen installed and running, to see how many spare mm I have.

 

[daddylongs]

So far so good - I ran Linx (all) for 20 passes and it maxes out at 52 degrees Cel. (This is at 44x100 = 4.4Ghz)

Speed Step does not seem to work so far but I am not concerned.

Does anyone have a good recommendation for a DVD ISO image I can boot to and run Linx, etc? The UBCD tests do not seem to push my computer very hard and misreport CPU speeds. Also the StressLinux distro is not user friendly at all.

Thanks!

 

[daddylongs]

Update on progress so far. You guys are good at those guesses so far 4.5x100 1.30v 65C


note the two different LinX runs - one with 4 threads and one with 8 - do those GFlops look reasonable for my speed? 2600k@4.5

I am installing an updated cooler tonight, so I will have updates on that later.

Edit: corrected image link

 

[Accord99]

With Windows 7 and SP1, the AVX enabled version of LinX should hit well over 110 GFlops:

http://www.xtremesystems.org/forums...-Stable-Club&p=4115565&viewfull=1#post4115565

 

[tweakboy]

It will take at least 1.45 - 1.5 to get 5g, unless you have a very rare chip ...

l1.45v and 1.50v is a big difference, You need supreme cooling not a no name HSF.

If I had a 2600k I would put 1.480v stable no drop @ 5Ghz . Can you imagine these new Intel chips will be 6 core 12 threads @ 5Ghz . should have you set for rest of your life hehe. Or until 2020.

 

[daddylongs]

l1.45v and 1.50v is a big difference, You need supreme cooling not a no name HSF.

If I had a 2600k I would put 1.480v stable no drop @ 5Ghz . Can you imagine these new Intel chips will be 6 core 12 threads @ 5Ghz . should have you set for rest of your life hehe. Or until 2020.

1.480v sounds dangerously high for a stable 24/7 system.


I am already stable at 4.7 (47x100) @ 1.375v. I will push it to 1.4v soon but I want to be careful and not break anything. For example, at these speeds I am getting up to 75C temps. I have heard before not to go above 70 for 24/7 stuff. Thoughts?

 

[Idontcare]

Update on progress so far. You guys are good at those guesses so far 4.5x100 1.30v 65C


note the two different LinX runs - one with 4 threads and one with 8 - do those GFlops look reasonable for my speed? 2600k@4.5

I am installing an updated cooler tonight, so I will have updates on that later.

Edit: corrected image link

Looks like you aren't running with AVX, your GFlops are way too low.

I don't know for sure, but I would assume the lower GFlops means less "work per unit time" is being done by the CPU and as such the power-consumption and CPU temperatures are also unintentionally lower than they would otherwise be if you were using AVX in LinX.

 

[daddylongs]

Looks like you aren't running with AVX, your GFlops are way too low.

I don't know for sure, but I would assume the lower GFlops means less "work per unit time" is being done by the CPU and as such the power-consumption and CPU temperatures are also unintentionally lower than they would otherwise be if you were using AVX in LinX.

My first take on the AVX usage in LinX was that it's an unnecessary, artificial benchmark. Are there any mainstream applications that use AVX?

I am already getting ~100Gflops in IBT, so that seems to be a good sign as a benchmark tool. No?

Regardless, for you, I'll do it; AVX inbound. ()

 

[Idontcare]

My first take on the AVX usage in LinX was that it's an unnecessary, artificial benchmark. Are there any mainstream applications that use AVX?

I am already getting ~100Gflops in IBT, so that seems to be a good sign as a benchmark tool. No?

Regardless, for you, I'll do it; AVX inbound. ()

It has nothing to do with relevance of the instruction set and more to do with the purpose of running IBT in the first place.

AVX should increase your peak temps during the stress test, which is the goal and purpose of running IBT in the first place.

But I am merely hazarding a guess that it actually does result in elevated temps, hence my questions on the subject.

For benchmarking you should look to other apps like cinebench, PCMark7, Fritz Chess, etc IMO.

 

[Dadofamunky]

I doubt he'd hit 127 GFlops anyway. I ran LinX and got... between 56 and 57 GFlops. Something is wrong with his box if he's getting those results. Maybe memory settings.

 

[daddylongs]

I doubt he'd hit 127 GFlops anyway. I ran LinX and got... between 56 and 57 GFlops. Something is wrong with his box if he's getting those results. Maybe memory settings.

I am a bit confused about what you think is worng...


I just did this experiment:

2600k @ 4.7GHz, 4GB RAM@1333, Win7x64_sp1

Can any of you show me how this scales as we get into larger RAM sizes? This is a pretty easy experiment to run. I only have 4GB though.

To me it looks asymptotic towards 125-ish. I could believe that I could get up to the 125-130 range if I push up the volts some more and start to actually push my RAM (DDR3-2000)

 

[Idontcare]

I am a bit confused about what you think is worng...


I just did this experiment:

2600k @ 4.7GHz, 4GB RAM@1333, Win7x64_sp1

Can any of you show me how this scales as we get into larger RAM sizes? This is a pretty easy experiment to run. I only have 4GB though.

To me it looks asymptotic towards 125-ish. I could believe that I could get up to the 125-130 range if I push up the volts some more and start to actually push my RAM (DDR3-2000)

I've got 16GB in my box, the numbers you saw in my graph above were for tests with ~14.3GB IIRC.

Why is your LinX lower than IBT? LinX is supposed to built on IBT, just different GUI. Are you sure you got the latest rev of LinX?

I'll run some lower-memory tests later, if no one beats me too it in the meantime.

I gotta ask though - $315 for a 2600K but not enough coin in the budget to get yourself >4GB? Are you running 32bit, is that why?

 

[daddylongs]

I've got 16GB in my box, the numbers you saw in my graph above were for tests with ~14.3GB IIRC.

Why is your LinX lower than IBT? LinX is supposed to built on IBT, just different GUI. Are you sure you got the latest rev of LinX?

I'll run some lower-memory tests later, if no one beats me too it in the meantime.

I gotta ask though - $315 for a 2600K but not enough coin in the budget to get yourself >4GB? Are you running 32bit, is that why?

I've got 16GB in my box, the numbers you saw in my graph above were for tests with ~14.3GB IIRC.

Why is your LinX lower than IBT? LinX is supposed to built on IBT, just different GUI. Are you sure you got the latest rev of LinX?

I'll run some lower-memory tests later, if no one beats me too it in the meantime.

I gotta ask though - $315 for a 2600K but not enough coin in the budget to get yourself >4GB? Are you running 32bit, is that why?

I am not sure why all my LinX scores are low compared to IBT. I have checked that both are the new versions (0.6.4 and 2.5.2), doubled checked by downloading new copies, got same results.

Both are basically GUI wrappers for the same tool right? I am not sure what diagnostic test to do next to understand why there is a difference.

What versions are you using?


(Point taken. More DDR3 is coming, so far I have been re-using the RAM from my last rig (Gigabyte 775 mobo had both DDR3 and DDR2 slots)

 

[Dadofamunky]

Heh. I'm the one who was confused. I never noticed that IBT/LinX scales with memory used for the test.

Just for the record, I get 61 Gigaflops and change running a 6.3 GB test. LinX 0.6.4. This tells me in general that synthetic benchmarks mean less than you think.

 

[daddylongs]

Have you tried running IBT 2.5.2? I have found it to more consistently run at predictable GFlops. LinX has sometimes without apparent reason slowed down into the 70-80 range.