Intel 4.5GHz & 5GHz LinX Stable Club

Printable View

Show 100 post(s) from this thread on one page

12-05-2010, 10:36 AM
Sniper

Quote:

Originally Posted by SimpleTECH

Sniper.... I am jealous...

Congrats. :up:

Thanks. Your 3845A935 seem to be about equal to my other W3520 (also 3845A935).

Quote:

Originally Posted by CryptiK

Really nice work there, I am envious of your boards max stable bclk. All these 220 - 230+ bclks, I can't go over 218 lol. One day I will have to get a better board and test mine in winter.

Thanks. I'v tried alot of X58 mb's, and always ends up back with this old GB X58 Extreme. UD7 rev 1 and a Classified are the only other boards that could (almost) match its bclk/stability. I also belive that the cpu is what limits max bclk, as long as the mb can "keep up" with the cpu.

Quote:

Originally Posted by Johnny87au

@ sniper whats ur 24/7 overclock?

Lately its been running at 217x21=4557 HT vcore = 1.38 load.

Quote:

Originally Posted by Sam_oslo

Good one :up:

Long time no see, good to see the old man hasn't given up yet, :rofl:

Dont have much time to play with hw any more, but still a lot of fun when I find some time :)

As for low GFLOPS, heres my experience with a gulf 3003Bxxx (sold now):
I had no problem running LinX at about 4.6 with it, but GFLOPS was low (40-50) and it was not stable at the same settings when testing with Prime. So I didnt enter it in this club as I figured there was something wrong with LinX and Gulfs.
12-05-2010, 11:16 AM
donmarkoni

Quote:

Originally Posted by some_one

Ummm, maybe donmarkoni would be kind enough to do 3 runs @ 4GHz of 64-bit with say 4GB of mem (24000). I would expect ~90% of full SSE with no HT (6 threads) and ~75% of full SSE with HT (12 threads).

HT = ~72GFlops
No HT = ~86GFlops

FWIW even though using SMT (HT) gives lower Flops it uses more power, ~10% more on my 860, so is probably more stressful than the higher Flops on separate physical cores.

No problem. :up: As soon as I get home, in about hour and a half, or two.
If I understand you, I should run 3 loops at 4GHz with 4GB HT OFF, then HT ON?
I have some s/s of previous 5 loop runs at lower then 4.5GHz, so I can post for comparison purposes.
12-05-2010, 03:36 PM
PiLsY

Been making the most of the freezing cold we've had here. Anyway, lynnfield time!

Board - Same old P55-UD5
Ram - G.skill Eco c7
CPU - i5 760 Batch info to follow - lost the bit of paper I wrote it on :lol:. Its an early one though.

CPU limit was 4445mhz on air with 8c ambient. Finally watercooled it (simple loop - Supreme LT, Magicool 360XT and EK DCP2.2) and took it out to the workshed for some -4c ambient :).

4.5ghz was quite easy so I decided to try and max out efficiency. Early runs were doing 58-59gflops, but with some tweaking and ramrunning 2050mhz 8-9-8-1T I managed to average over 60.5gflops. Pretty impressive :). Had to run with all eist and c1e off as single core turbo mode was kicking in and boosting it up to 25x at idle. Need to move to 21x multi to move higher on cpu speed, this cpu clocks better with eist enabled. Problem is that with this cpu I cant get the bclk over 225. 226 boots but isnt stable, 227 wont post. Swap to the 530 and I get 238, with the 550 it can do 244. Definitely a cpu limit. Anyway, 4.5ghz high efficiency screenshot below:

http://i169.photobucket.com/albums/u...rdllcwater.jpg

So with that said lets push the cpu a bit :). Ram down to 8x and some bclk testing got me to 1.58vddr, 1.38vtt 6-8-6-1T @ 225bclk / 1800mhz. Very good result - the clarkdales cant run these same eco sticks at cas 6 at all.

And yes you guessed it...thats 21 x 225 for 4725mhz :). Second screenshot shows voltage under load - monstrous vdroop on this P55-UD5. Running standard LLC - level 1 and level 2 limit the overclock and give me more heat from the board. Just cranking the volts to counteract the drop. Still its getting close to 0.15v idle to load....

Anyway, here's the money shot :lol:.

http://i169.photobucket.com/albums/u...604725linx.jpg

And the voltage under load (and memory timings for good measure)....

http://i169.photobucket.com/albums/u...21x225load.jpg

First lynnfield in the club?

[Edit : tried cropping all the pics a bit but the forum seems to cache pics of the same name - if anyone objects about the size ill shrink the other two a bit tomorrow!]
12-05-2010, 03:56 PM
PiLsY

Sniper - damn nice job dude, got my 4.811 beat :up:. Looks like im going to have to be lugging the i7 into the workshed to get back in the top 3. I may be able to squeeze another 1-2bclk out of it. Hope you dont have any backup scores cos that really is the utter limit for this board/cpu/ram combo lol.
12-05-2010, 04:36 PM
zsamz_

Quote:

Originally Posted by PiLsY

Sniper - damn nice job dude, got my 4.811 beat :up:. Looks like im going to have to be lugging the i7 into the workshed to get back in the top 3. I may be able to squeeze another 1-2bclk out of it. Hope you dont have any backup scores cos that really is the utter limit for this board/cpu/ram combo lol.

i just got an ek supreme hf i'm close to gettin 231x21:yepp:
my heatkiller just not cutiin it
as soon as i hit 85c:explode:
so far i'm only good for 5 min of linx @231x21:(
12-05-2010, 04:56 PM
donmarkoni

Quote:

Originally Posted by some_one

Ummm, maybe donmarkoni would be kind enough to do 3 runs @ 4GHz of 64-bit with say 4GB of mem (24000). I would expect ~90% of full SSE with no HT (6 threads) and ~75% of full SSE with HT (12 threads).

HT = ~72GFlops
No HT = ~86GFlops

FWIW even though using SMT (HT) gives lower Flops it uses more power, ~10% more on my 860, so is probably more stressful than the higher Flops on separate physical cores.

I hope I understood you right, so here it goes (actually, added UnCore variations):
1-LinX_HT_4000-3200-2000_77.65_GFLOPS
2-LinX_HT_4000-3600-2000_78.99_GFLOPS
3-LinX_HT_4000-4000-2000_79.82_GFLOPS
4-LinX____4000-3200-2000_87.03_GFLOPS
5-LinX____4000-3600-2000_87.64_GFLOPS
6-LinX____4000-4000-2000_88.74_GFLOPS

1.http://img809.imageshack.us/img809/4...2002000.th.jpg 2.http://img291.imageshack.us/img291/2...6002000.th.jpg 3.http://img689.imageshack.us/img689/5...0002000.th.jpg 4.http://img684.imageshack.us/img684/6...2002000.th.jpg 5.http://img291.imageshack.us/img291/7...6002000.th.jpg 6.http://img97.imageshack.us/img97/474...0002000.th.jpg
12-05-2010, 05:21 PM
miahallen

Some of you guys have WAY too much time on your hands :rofl:
12-05-2010, 05:33 PM
OC Nub

Nice job PiLsY. 3rd lynnfield in the club, but you are the highest.
12-05-2010, 05:33 PM
donmarkoni

Quote:

Originally Posted by miahallen

Some of you guys have WAY too much time on your hands :rofl:

I'm sorry for taking your time, busy man, but you really don't need to read all this if you have work to do.
And yes, I will always find time for my hobby and if someone asks me, will try to help. :cool:
12-05-2010, 06:05 PM
Sniper

No backups PiLsY, so its all up to you (and zsamz) :up:
12-05-2010, 06:56 PM
miahallen

Quote:

Originally Posted by donmarkoni

I'm sorry for taking your time, busy man, but you really don't need to read all this if you have work to do.
And yes, I will always find time for my hobby and if someone asks me, will try to help. :cool:

Not at all....I think it's awesome :D Great work with the Xeon BTW :up:
12-05-2010, 07:25 PM
zsamz_

Quote:

Originally Posted by miahallen

Some of you guys have WAY too much time on your hands :rofl:

not really lol
it took me a many months just to find the right board and to find the right cooling solution
i wont give up lol
12-05-2010, 08:28 PM
OC Nub

3x1gb kit loose timings, MSI XPower, Water with some AC. Very fast run but at this vcore using water I dont think I will try my 6gb kit, would take 1hr. 1.6+vcore :shocked:

This is no golden batch chip, but for 179.00 it was a great deal.
12-05-2010, 08:51 PM
CobraXP

http://img714.imageshack.us/img714/6692/46noeoors2.jpg
Voltage 1.44v. Cooling on my sig.
12-06-2010, 12:50 AM
PiLsY

Epic fail last night with the 950 - condensation issues. All ok, but had to quit early. 4.833 is definitely there for the taking if I can sort the condensation on the cpu block baseplate out. Worried itll soak into the cpu socket. -5c lowest temp outside last night and real bad freezing fog - it was snowing ice crystals!
12-06-2010, 01:51 AM
Johnny87au

@oc nub, 1.6vcore... dare u to bench n super PI it :D
12-06-2010, 01:53 AM
Sniper

Wow, great run OC_Nub! Somehow I got the feeling I'v got to sit and freeze another night very soon. :)
12-06-2010, 04:25 AM
OC Nub

Thanks.
You guys did much longer runs using 6gb kits so this really doesnt compare to an hour run, but at 1.6v this is all Im willing to try.
12-06-2010, 09:26 AM
some_one

Quote:

Originally Posted by donmarkoni

I hope I understood you right, so here it goes (actually, added UnCore variations):
1-LinX_HT_4000-3200-2000_77.65_GFLOPS
2-LinX_HT_4000-3600-2000_78.99_GFLOPS
3-LinX_HT_4000-4000-2000_79.82_GFLOPS
4-LinX____4000-3200-2000_87.03_GFLOPS
5-LinX____4000-3600-2000_87.64_GFLOPS
6-LinX____4000-4000-2000_88.74_GFLOPS

Thanks donmarkoni, that pretty much ties in with those ballpark percentages, well you score a bit higher but I think some variation would be usual. Nice that you have an unlocked uncore. What do you think, maybe use that formula or something like it to give people some idea of what to expect?

EDIT: Clarkdale (i5-655k) seems to work differently.

Gautam, I don't know why there are so may low GFlops. What if you try with less memory, maybe using "all" hits the OSes resources too much. Even running 4GHz and a problem size of 6000 should hopefully give some results in the 60GFlops range with 12 threads and only take a few seconds to do 3 runs. Might also be worth checking the taskmanager to see if the CPU is being fully utilized.
12-06-2010, 11:57 AM
donmarkoni

Quote:

Originally Posted by some_one

Thanks donmarkoni, that pretty much ties in with those ballpark percentages, well you score a bit higher but I think some variation would be usual. Nice that you have an unlocked uncore. What do you think, maybe use that formula or something like it to give people some idea of what to expect?

EDIT: Clarkdale (i5-655k) seems to work differently.

Gautam, I don't know why there are so may low GFlops. What if you try with less memory, maybe using "all" hits the OSes resources too much. Even running 4GHz and a problem size of 6000 should hopefully give some results in the 60GFlops range with 12 threads and only take a few seconds to do 3 runs. Might also be worth checking the taskmanager to see if the CPU is being fully utilized.

Hmm... IDK about the more accurate formula, cause there are just too many variables, but in How to run LinPack stress test(LinX/IBT)properly-an explanation(maybe a guide) is pretty much explained everything that can impact the performance.

I was thinking the same today, while working :D , about the "all" button and the issues it could make when asking all memory from OS, which can start using PageFile and impair the performance. Then I checked a couple of results here and found out that "all" was not in common use for low GFLOPS runs. :shrug:
12-06-2010, 01:14 PM
zsamz_

nice run there Ocnub:up:
12-06-2010, 02:09 PM
OC Nub

Thank you zsams.

Keep failing 231x21, have the ac on and window open with a fan blowing the cool air in.
12-06-2010, 02:46 PM
zsamz_

Quote:

Originally Posted by OC Nub

Thank you zsams.

Keep failing 231x21, have the ac on and window open with a fan blowing the cool air in.

mine just shut down overheated:rofl:
i messed up on block install:mad:
heatkillers are just a pain to swap cpus so i bought a supreme hf:D
man this thread hurtin my pocket lol
12-06-2010, 04:18 PM
OC Nub

^Agreed, put a dent in my pocket also, and my 920 isnt to happy.
12-07-2010, 05:48 AM
Dahaka

Quote:

Originally Posted by OC Nub

^Agreed, put a dent in my pocket also, and my 920 isnt to happy.

Hi men, wich revision are you using on Xpower?. I have 2 deads Xpower over here and i don't want to try another one to get the same result.

I got a ES and have the problem with qpi, uncore and high memory frecuency :(.

Show 100 post(s) from this thread on one page