Intel 4.5GHz & 5GHz LinX Stable Club

[zoson]

Toasty, but performing like a REAL 4.5GHz cpu should. All these people under 90GFlops should try again.

Core i7 970 [6C 6T] @ 4512.4 MHz | 1.51500 v. | ASUS Rampage III Extreme | Water (EK Supreme HF FG) | 3016B246

[kitfit1]

All these people under 90GFlops should try again

Getting over 90GFlops is dead easy if you are running with HT off zoson, as you have found out. But try running the same clock with HT on, you will then find a completely different result.

[zoson]

Getting over 90GFlops is dead easy if you are running with HT off zoson, as you have found out. But try running the same clock with HT on, you will then find a completely different result.

I was not talking about the guys with HT on. Although at 4.3 with HT I'm over 86GFlops, so I can imagine being near or above 90 at 4.5 with HT on.

The problem I was talking about earlier was the large memory size HT issue that apparently exists.

[R37ribution]

Toasty, but performing like a REAL 4.5GHz cpu should. All these people under 90GFlops should try again.

Core i7 970 [6C 6T] @ 4512.4 MHz | 1.52500 v. | ASUS Rampage III Extreme | Water (EK Supreme HF FG) | 3016B246

Were there any voltages and/or settings that were key in improving the Gflops for your processor? I start to loose performance in this respect right after 205bclk or it is intermittent and I have to try restarting LinX several times. I've messed with CPU PLL, IOH, vtt, memory timings and subtimings, round trip latency, DRAM ref volts, CPU/IOH clock skew, CPU diff amplitude and can't get consistant performance. I've also disabled a ton of windows services like superfetch to eliminate any program in the background.

[LASeR10]

LASeR10 | Core i5 2500K [4C 4T] @ 5405.6MHz | 1.528 | Asus P8P67 Pro | Water | L051B314-xxxx
Submission #04

[R37ribution]

LASeR10 | Core i5 2500K [4C 4T] @ 5405.6MHz | 1.528 | Asus P8P67 Pro | Water | L051B314-xxxx
Submission #04

Holy crap! Why so many? You tryin to get all those in,the club?

[l0ud_sil3nc3]

Holy crap! Why so many? You tryin to get all those in,the club?

nope looks like he is just binning to find the golden willy wonka bar

[LASeR10]

hmm L051B314 for my buddy...

[zalbard]

Going through all the entries is taking longer than anticipated. Two pages to go. Should be done soon.

[l0ud_sil3nc3]

Going through all the entries is taking longer than anticipated. Two pages to go. Should be done soon.

once again just wanted to say thanks for maintaining this thread, keep it up

[zalbard]

Okay, the first post is up to date.
I have also updated the link for the up-to-date LinX distributive.
All the submissions that have made it into the OP are (in one way or another) outstanding. Good job, everyone!
I will only comment on the selected few just so the post isn't a mile long.

Can I join? This is after owning an i7-965 and struggled to get 4 GHz on air.

xarot - i7 990X [6C 6T] @ 4500.4 MHz - 1.368 V - Asus P6T6 WS Revolution - AIR - ES

Yes, you can! Very nice chip, and welcome!

Newer Linpack, 10.3.3.006.

Nice find, thanks.
Looks like it works with AVX better, and people won't have to run LinX on 4 threads with HT on any more.

Updated run - screen caps showing correct (lower) vcore and ran with larger problem size using latest linpack binaries.

990X | 4505 MHz | 1.336v | 3550 MHz uncore | 1.244 Vtt | DDR3-1638 MHz | 7-7-7-20 1T | 1.60 vdimm

Glad that 990X is treating you well! Good job tweaking it, also!

Update 3:
Can I be the first Gulftown to Join the: 5GHz Club <Over 100 GFlops> (Sandy Bridge and others)……

needforspeed| Core i7 990X [6C 6T] @ 5002.9MHz | 1.640 | ASUS Rampage III Extreme | Water | 3045B094

I have my eye on 5100Mhz.. but don’t know if I have the rocks.. to take the Vcore any higher.. although the temps are fantastic..

I've only been gone for two weeks and people are already breaking into restricted areas!
That's absolutely fantastic, man! Congrats, and well done!

Testing Kingston 2250CL9 kit... Vcore can be lower, have to keep VTT around 1.4 with this crap CPU...

Leeghoofd| Core i7 990X [6C 6T] @ 4636MHz | 1.42 | ASUS Rampage III Formula| Water | 3045A596

Just 4.6GHz... Pff, what are you even doing here?!
"Crap" CPU, indeed, especially the IMC.
I think LinX isn't the best way to test your RAM, though, but you know that. Looks good nonetheless!

Core i5 2500k [4C 4T] @ 5413.1 MHz | 1.610 v. | ASUS P8P67 PRO | Water (EK Supreme HF) | L051B314

These are some crazy clocks and voltages! I hope your CPU is a durable one!
Grats on taking the top in your category!

LASeR10 | Core i5 2500K [4C 4T] @ 5405.6MHz | 1.528 | Asus P8P67 Pro | Water | L051B314-xxxx
Submission #04

Stiff competition! #2! Much nicer vCore, though!

Guys just a pointer and I hate to sound 'harsh', but this isn't a general oveclocking and discussion thread. If you don't have an OC of 4.5GHz or over validated with 20+ loops of LinX, and instead you just have general discussion of GFLOPS and 2 and 4 loop runs of LinX posts, best to post somewhere suitable for general stuff like that like your motherboard thread etc. I don't feel it's appropriate to expect people (and Zalbard) to wade through that stuff in here - save posting here for when you have got your settings down and worked hard to get that 20 loop run done.

A question here or there I think is fine, but I just hope this thread stays specific and on task rather than drifting. Most threads like this get forgotten about and never updated, Zalbard does an awesome job and I think it's the spirit of this thread to stay specific about 4.5+/20+ loops.

EDIT - I dunno, reading over my post I seem like some whinging old guy. Im all for helping each other out and stuff, just, I dunno it sounded ok in my head..

Well, I am not so sure. Some on-topic chatter is fine, IMO... This is a Club with members, not just a database. Makes the thread feel more 'humane' and 'alive'.
No need to have lengthy discussions about which alcohol you prefer or which video card gets the most 3DMarks, though. Just don't go off-topic too far.

once again just wanted to say thanks for maintaining this thread, keep it up

Thanks for the kind words, and you're welcome!
Glad you guys are having the fun.
Now that things are back on track and I have some free time I will think about updating the thread's formatting. Should look a lot nicer. Should be also able to get the results in within 1-2days, too. I actually keep thinking I need some software for simplified result input, lol.

[zoson]

Hi Zalbard,
Thanks for keeping up the thread. It's a really nice way to organize the community of extreme overclockers.

Might I suggest for the thread update that you sort by GFlops, rather than speed? It would simplify the maintenance of the thread by making it possible to remove the different sections for specific CPU's. It would also express the ranking as a performance level, rather than just a benchmark. The speed milestone would still be the barrier to entry, but competition would be based on performance!

It would also make it very easy to see the clock per clock comparisons between old and new generation cpu's. 'oh hey at 100Gflops, xx cpu is lower clockspeed'. The community could zero in on those people who know tricks that boost performance(to get their insight) without having to sort through every result to find who is actually performing well, vs who is just running at a high clock.

Thanks again!

[tool_462]

LASeR10 | Core i5 2500K [4C 4T] @ 5405.6MHz | 1.528 | Asus P8P67 Pro | Water | L051B314-xxxx
Submission #04

Very nice chip! Would like to see what it can do with proper AVX LinX installed

Since my 2500K at 4.8 does 130+ GFlops with AVX.

[l0ud_sil3nc3]

Very nice chip! Would like to see what it can do with proper AVX LinX installed

Since my 2500K at 4.8 does 130+ GFlops with AVX.

yup my 2500k at 4.8 maxes around 139.xx with SP1 installed, quite a difference from before

[skuldarin]

UPDATED:

CPU - 4.672ghz (133x35) @ 1.53125v | MEM - 1866 @ 9-9-9-24-82 2T (1.68275v) 12gb | UNCORE - 3733mhz (VTT 1.36875)
BATCH - 3013A540 | MOBO - ASUS R3E | COOLING - Water | 6C/6T | 101.2 Gflops average


wonder if this starts a "sub 5ghz, over 100gflops" group?

[newhit]

Managed to squeeze some more out of my chip
L042B076
water
Gigabyte GA-P67A-UD7 (B2)
Win7 with AVX (SP1)

[ac001]

ac001 | Core i7 990X [6C 6T] @ 4539MHz | 1.429 | ASUS Rampage III Black Edition | Water | -



regards
ac001

[zalbard]

Hi Zalbard,
Thanks for keeping up the thread. It's a really nice way to organize the community of extreme overclockers.

Might I suggest for the thread update that you sort by GFlops, rather than speed? It would simplify the maintenance of the thread by making it possible to remove the different sections for specific CPU's. It would also express the ranking as a performance level, rather than just a benchmark. The speed milestone would still be the barrier to entry, but competition would be based on performance!

It would also make it very easy to see the clock per clock comparisons between old and new generation cpu's. 'oh hey at 100Gflops, xx cpu is lower clockspeed'. The community could zero in on those people who know tricks that boost performance(to get their insight) without having to sort through every result to find who is actually performing well, vs who is just running at a high clock.

Thanks again!

Hi, and thanks.
The problem about this is, these GFlops are not really relevant to real-life performance. For example, currently non-HT CPUs produce more GFlops than those with HT enabled. Yet if you try some rendering software, HT-enabled CPUs will be way ahead.
Another issue is memory performance. Linpack seems to be quite sensitive to that. It doesn't really translate into real-world advantage, though. For example, a 4GHz CPU with 2GHz RAM may be faster in LinX than a 4.5GHz CPU with 1GHz RAM, but it will be the other way around in pretty much all of the applications.
Not to mention that produced GFlops are often inconsistent.
I think it makes sense the way it is now. LinX is used to measure stability, not performance.
However, it's up to the members to decide. If I see many more people supporting your motion, I can rearrange everything.

[l0ud_sil3nc3]

Hi, and thanks.
The problem about this is, these GFlops are not really relevant to real-life performance. For example, currently non-HT CPUs produce more GFlops than those with HT enabled. Yet if you try some rendering software, HT-enabled CPUs will be way ahead.
Another issue is memory performance. Linpack seems to be quite sensitive to that. It doesn't really translate into real-world advantage, though. For example, a 4GHz CPU with 2GHz RAM may be faster in LinX than a 4.5GHz CPU with 1GHz RAM, but it will be the other way around in pretty much all of the applications.
Not to mention that produced GFlops are often inconsistent.
I think it makes sense the way it is now. LinX is used to measure stability, not performance.
However, it's up to the members to decide. If I see many more people supporting your motion, I can rearrange everything.

I am going to have to pass as using gflops as a metric and as you were saying the numbers are inconsistent, and what really does matter is real world applications, I feel the current methodology is inline

[needforspeed]

I am going to have to pass as using gflops as a metric and as you were saying the numbers are inconsistent, and what really does matter is real world applications, I feel the current methodology is inline

I second that....

[zoson]

The problem is that the reason why gflops are inconsistent, is because of instability in the person's system. There is a reasonable expectation for raw calculations at a given speed. I have never seen my Gflops vary if the speed is kept the same.

I also take issue to the memory remarks... I've tested all the way from 1333 cas5 to 2200 cas10. The difference in gflops when cpu speed is kept the same is very small, as long as the memory size/problem set size is consistent. 1-1.5Gflops tops.

Also, if you don't use enough memory, you cannot stress the cpu correctly! The problem set will not be large enough to engage all the execution units of the processor!!! That is why there is lower Gflops with lower memory. The stress test is not complete. It is also why we see lower temps with lower memory sizes, and also why we see lower temps if HT is on and memory size is > 9gb... It is not difficult at all!!! I can complete linx 21 loops at 4.9ghz with 512mb memory size! Gflops is less, but it completes! Is that stable? no!!!!

In summary, I think the reason for inconsistent gflops is they are doing it wrong!!!!!

[l0ud_sil3nc3]

The problem is that the reason why gflops are inconsistent, is because of instability in the person's system. There is a reasonable expectation for raw calculations at a given speed. I have never seen my Gflops vary if the speed is kept the same.

I also take issue to the memory remarks... I've tested all the way from 1333 cas5 to 2200 cas10. The difference in gflops when cpu speed is kept the same is very small, as long as the memory size/problem set size is consistent. 1-1.5Gflops tops.

Also, if you don't use enough memory, you cannot stress the cpu correctly! The problem set will not be large enough to engage all the execution units of the processor!!! That is why there is lower Gflops with lower memory. The stress test is not complete. It is also why we see lower temps with lower memory sizes, and also why we see lower temps if HT is on and memory size is > 9gb... It is not difficult at all!!! I can complete linx 21 loops at 4.9ghz with 512mb memory size! Gflops is less, but it completes! Is that stable? no!!!!

In summary, I think the reason for inconsistent gflops is they are doing it wrong!!!!!

when I say the gflops vary I mean by tenths of a unit, and to truly test stability you can not just use one program, but rather a plethora of different ones, as well as the applications you will use on a day to day basis. I do however agree that it is easier to pass with a lower problem size, no doubt but as long as you are using at least 4GB I think you will be good.

well at the end of the day this is zalbard's rodeo so we will see what he wants to do, but I like having the list sorted by raw clockspeed

[some_one]

Hi Zalbard,
Thanks for keeping up the thread. It's a really nice way to organize the community of extreme overclockers.

Might I suggest for the thread update that you sort by GFlops, rather than speed? It would simplify the maintenance of the thread by making it possible to remove the different sections for specific CPU's. It would also express the ranking as a performance level, rather than just a benchmark. The speed milestone would still be the barrier to entry, but competition would be based on performance!

It would also make it very easy to see the clock per clock comparisons between old and new generation cpu's. 'oh hey at 100Gflops, xx cpu is lower clockspeed'. The community could zero in on those people who know tricks that boost performance(to get their insight) without having to sort through every result to find who is actually performing well, vs who is just running at a high clock.

Thanks again!

IMHO this thread is about running at a certain clock speed while verifying stability with Linpack. If you really want a performance comparison of FLOP performance throughput it might be a good idea to start a new thread with it's own rules to achieve that. You'll have to decide what you want, max FLOPS overall, FLOPS per core, FLOPS per GHz or maybe even FLOPS per watt for efficiency (may be hard to achieve) etc. Perhaps also some distinction between FPU, SSE, AVX.

[newhit]

Iam sure that "Zoson" has good scientific reasons for what he says but ppl like myself, and I'm sure many others, do not have the time (work, family) to go thru what is necessary to produce the required "Flops" stability.

I have an SS that I hope to get running soon and this thread gives me the ideal MHz targets to try to aim for without taking me to areas of o/c'ing that require too deep a knowledge of the wherewithal that a casual ocker like myself wont bother with. When I have problems with my setup I know there are plenty of knowledgeable ppl in this thread to answer my (hopefully) not too difficult questions, thus allowing me to enjoy the thread for what it is................

a challenging bit of fun

OOPS! I hope my recently posted "challenging bit of fun" gets in Mr Zalbard .

[hellper]

Hi everybody!

A little bit old-fashioned, I know, but here is my application to the 4.5+ GHz club:



(Linx 20+ 64-bit maximum, E8500, 4503mhz, 1.36v, Gigabyte P35-DS4, Thermaltake Big Typhoon aircooling, 15 degrees Celsius ambient, OS Win7 SP1 64-bit in diagnostic mode)

The run finished successfully, in fact (all 25 iterations), and shortly after that I reran the test (also with success), just to make sure everything is OK; don"t want to waste space for the pictures, but they are available on demand.

(Gigabyte boards are Linx- and IBT-friendly, I guess.)