PCMark Vantage World Records

[onex]

i'm thinking +2k for 5.5 maybe even higher,
and wonders how much Corsair setup do on stock,
u did get quite a jump, apparently from 30.3K @4.65 to 31.3 @4.75...
33K on 4.8 (yet prob. with some side tweaking).

you got this:


on SS, 1.752, your chip afai can tell, will take ~the same amount of voltage for this kind of OC.
i don't know if you'll go there, yet by earlier results, it seems, on 5.5,
you might be able to get 38K or even scratch the 40.
-that's an estimate-.

[onex]

Mike, i think steve is doing a great effort here,
no one gets to be an expert unless he's going through some harsh time at first.
maybe instead of banging into people, share u'r knowledge, we would all benefit from this.

E:
talked over PM.

[SteveRo]

i'm thinking +2k for 5.5 maybe even higher,
and wonders how much Corsair setup do on stock,
u did get quite a jump, apparently from 30.3K @4.65 to 31.3 @4.75...
33K on 4.8 (yet prob. with some side tweaking).
i don't know if you'll go there, yet by earlier results, it seems, on 5.5,
you might be able to get 38K or even scratch the 40.
-that's an estimate-.

Wow, Onex, nice numbers - excellent clock, one of the best I have seen!
Under SS how long could I get away with puting 1.7+v through it - could I do that for an hour or two under load w/o breaking anything?

[onex]

this isn't my chip friend, this is taken from HWbot,
you'r chip was going ~1.584@ 4.8GHz where chips alike with these clocks at this speed, went fine ~1.72 @ 5.5GHz on SS and ~the same proportional voltage,
you can have a look at the CPU-Z benchmark results there, they scale from ~4.5 or so and up to 6.8 by NickShin here from the forum,
most guys till 5Ghz still use air cooling to reach these clocks, some uses H2O and LN2 to reach higher,
i brought the SS specifically to show that 5.5 is doable with a regular 980 at ~1.75+-.
some of the chips there where going much lower, these were generally ES chips going even IIRC ~1.6 (talking INSANE numbers for 5.5Ghz 980 chip),
NickShin has brought his CPU, up to 1.9+ core voltage,
what i'm saying, is that these chips CAN generally handle 1.7+ core voltage without any signficant effort,
they can go up to 1.9 and still remain healthy,
for a hour+ run, i'll keep a good eye on temperatures, i think it is doable, and not too dangerous,
it seems to be still on the safe scale for extreme OC'ing.

afai can tell, u'r CPU won't be even under 100% load throughout the all test.
is that correct?

[SteveRo]

With task manager open and watching - at 4.8 - seems like the highest I saw was about 80% utilization

[onex]

yeah, that was exactly what u'v said last time when members here asked you,
iirc, it was between 30-80%,
i don't think, an hour+ of testing 30-80% CPU usage would cause any damage,
as a matter of fact, i don't think even 6Ghz would cause any serious damage at these loads,
these CPU's are generally capable of working up to ~10 years running stock,
they can probably go ~2-3 years 4.5, a year+ working 5.0Ghz and we're talking here 100% usage DC working or probably even priming for few month's...
there was a guy (i think it was DAK) from the WCG team here who said long ago his 920 has lost some of his cache after going (maybe 4Ghz) for few month's,
that's the only thing that came into my mind when u showed u'r worries over any damage happening to u'r OC'ed 980,
i think an hour is very minor, yet maybe u should ask there about they're observations..
it could've been a semi-bad chip too,

generally, i would've go for it without too many worries.

[SteveRo]

Thanks Onex.

[Expat GriZ]

Steve...I'd like to preach a little bit of caution here. There are no set rules when using high V & VTT...some chips can take it, some die without warning rather quickly. IF your SS can hold 6 cores @ -35+ for this venture....I think that would be about max. So whatever speed you can get to maintain these temps. Forget about 6ghz, not going to happen. I think 5.2 would be about max....these b*ggers run hot and with this being such a long Benchmark..time will be against you as well. You just won't be cold enough to get really big clocks out of your chip. You would need to at least hit -90 for 5.5 etc. Sorry, just my opinion but better to be safe than sorry and I don't want to see you kill your chip.

[SteveRo]

Expat, yep I plan to start slow and expectations are adjusted accordingly.
Hopefully by this time next week I will have more to report, thanks for the help.

[Anvil]

Looks like your going to need the new cooling Steve.

Link

Awesome comeback from Mike
Great fun watching the scores increase steadily.

[SteveRo]

Excellent work Mike, congrats.
Wow I think I got to keep the WR all of 48 hrs this time!

correction - actually 54 hrs!

[Computurd]

This is getting to be somewhat like watching tennis....

[SteveRo]

I should be able to post a new score in another week or so.
See - http://www.xtremesystems.org/forums/...5&postcount=11
Very pumped!

[onex]

what is this 300W thing?
is it power consumption or (seems more like) heat draw capability?

so basically, this setup would be able to "inhale" a CPU running 300W and keep it down to -39C?

[Expat GriZ]

what is this 300W thing?
is it power consumption or (seems more like) heat draw capability?

so basically, this setup would be able to "inhale" a CPU running 300W and keep it down to -39C?

yup.....tested to hold a load of 300 watts @ -39C.

[onex]

expat, why did u say u think 5.5Ghz OC would have to go -90?
i'm asking this as normal OC goes fine on room temperature, and up to the specific processor TM limit,
why does higher clocks require (by what u say) such low temperatures?
i've heard about cold-bugs cold-boots etc. yet why does a cpu require for exp. -130 and lower to reach higher then 6+ Ghz?
is there any point when such temp's becomes a necessity ?.

[onex]

expat, why did u say u think 5.5Ghz OC would have to go -90?
i'm asking this as normal OC goes fine on room temperature, and up to the specific processor TM limit,
why does higher clocks require (by what u say) such low temperatures?
i've heard about cold-bugs cold-boots etc. yet why does a cpu require for exp. -130 and lower to reach higher then 6+ Ghz?
is there any point when such temp's becomes a necessity ?.

[Jor3lBR]

expat, why did u say u think 5.5Ghz OC would have to go -90?
i'm asking this as normal OC goes fine on room temperature, and up to the specific processor TM limit,
why does higher clocks require (by what u say) such low temperatures?
i've heard about cold-bugs cold-boots etc. yet why does a cpu require for exp. -130 and lower to reach higher then 6+ Ghz?
is there any point when such temp's becomes a necessity ?.

The new intel cpu's scale with cold not with volts.

My i5 670 maxed out around 5.6Ghz (HT on) with -101oC under minimal PCM05 load.

Same cpu with same volts runs around 6.5ghz with -196oC. That's more then 1000Mhz scaling with temperature.

[onex]

hmm, i see,
this is quite shocking,
i wonder whether the same chip would run faster cooler.
i've seen these I5's reach 7Ghz here on the forum,
this might be bad news for steve, maybe it would be hard getting the 980 much further from the 4.8 mark on air.
i've heard electronic cicuit runs better on cold temperature, and see now why this 'kid' from the sub-zero cooling section here at the forum is building a 3.5-4.5KW cooling solution,
if a 920 running 4.5 can draw ~300-350? W then i'm not sure if a 980 (stock 3.33) would be able to reach 5 so easily with a 300W cooling mechanism,
taking u'r chip into account, a 3.46Ghz which has to have -101 to reach 5.6, i feel like a fool saying what i did few posts back,
computers are never linear, and thinking this way only comes when u'r unaware of other implications/complications,
this kind of thinking brings u into making many mistakes.

it's never the bigger, the faster, the better, it's all about the small things,
i'm wondering whether there is a way to see things, or approach things differently.

[SteveRo]

@ Jor3IBR - I guess I should be able to get 5 maybe 5.2 at -39C?
Where should I start with vcore - around 1.7v, maybe 1.68v?
What should I run cpu/vtt at - 1.4v or higher?
cpupll I have been using 1.9v

[mbreslin]

I should be able to post a new score in another week or so.
See - http://www.xtremesystems.org/forums/...5&postcount=11
Very pumped!

I always did hate your guts steve.

[Planet]

hmm, i see,
this is quite shocking,
i wonder whether the same chip would run faster cooler.
i've seen these I5's reach 7Ghz here on the forum,
this might be bad news for steve, maybe it would be hard getting the 980 much further from the 4.8 mark on air.
i've heard electronic cicuit runs better on cold temperature, and see now why this 'kid' from the sub-zero cooling section here at the forum is building a 3.5-4.5KW cooling solution,
if a 920 running 4.5 can draw ~300-350? W then i'm not sure if a 980 (stock 3.33) would be able to reach 5 so easily with a 300W cooling mechanism,
taking u'r chip into account, a 3.46Ghz which has to have -101 to reach 5.6, i feel like a fool saying what i did few posts back,
computers are never linear, and thinking this way only comes when u'r unaware of other implications/complications,
this kind of thinking brings u into making many mistakes.

it's never the bigger, the faster, the better, it's all about the small things,
i'm wondering whether there is a way to see things, or approach things differently.

The 32nm dual cores hitting 7GHz wouldn't do much for this benchmark as the would probably need to be around 6.3-6.5GHz stable to run the bench. Where as I can just run my 980x at 5.9GHz+ and easily beat the dual core at 6.5GHz. As far as what Steve will hit on a single stage really depends on the chip some chips scale some don't. I had a few chips on single stage that only ran 5.3-5.4.

[onex]

I had a few chips on single stage that only ran 5.3-5.4.

o.k, that's actually quite good ,
i was never expecting SS to pass 5.5+- actually,
we can really make a lot of testing to see the chip behavior with voltage vs temperature and Ghz.
this should be very interesting, actually, the real meaning of overclocking.

[Jor3lBR]

@ Jor3IBR - I guess I should be able to get 5 maybe 5.2 at -39C?
Where should I start with vcore - around 1.7v, maybe 1.68v?
What should I run cpu/vtt at - 1.4v or higher?
cpupll I have been using 1.9v

First thing to do is see how much it scaled with the better cooling.

Use same volts you are running currently on your chiller and slowly increase FSB. Run PCMV CPU tests and see until when it will still pass.

With that you gonna have a good picture on how is the scaling on your cpu.

After that I would start increasing vcore and a tad of your vtt, pll can stay in that area 1.9 to 2.

You will reach a limit there when the SS won't be sacling anymore with volts because the heat will peak quickly during PCMV cpu tests, and thats when you will notice that none of the volts will help you with fsb.

PS.: You have a UD7 did you use the mod hicookie suggested?

[SteveRo]

First thing to do is see how much it scaled with the better cooling.

Use same volts you are running currently on your chiller and slowly increase FSB. Run PCMV CPU tests and see until when it will still pass.

With that you gonna have a good picture on how is the scaling on your cpu.

After that I would start increasing vcore and a tad of your vtt, pll can stay in that area 1.9 to 2.

You will reach a limit there when the SS won't be sacling anymore with volts because the heat will peak quickly during PCMV cpu tests, and thats when you will notice that none of the volts will help you with fsb.

PS.: You have a UD7 did you use the mod hicookie suggested?

Very good, sounds like a sound plan.
Hopefully this coming weekend I can get it all together.
Yes, I saw the hicookie mod - I thought that was only for deep subzero - like -100 stuff.