Skylake hits 7ghz in test

[Rakehellion]

What a totally pointless exercise. If you can't run the whole chip, then I'd throw out the results.

I think the fact that it runs on liquid nitrogen is the part that makes it pointless.

AMD is still cleaning their clock on these insane overclocks -- An FX-8370 was running all 8 cores at 8.722 Ghz.

http://wccftech.com/stlit-hits-world-record-amd-fx-8370-clocked-87ghz/

Why are AMD chips better at overclocking? I know many of them run at higher stock speeds and probably have lower instructions per clock, but why are they able to clock higher?

 

[DrMrLordX]

Bulldozer/Piledriver have long pipelines, which tends to make for high operating frequency (it was the same deal with Netburst, more-or-less). Or at least, that was my impression.

 

[dark zero]

No disrespect to Lam, but Toppc arguably did a better job under LN2 anyway:

http://www.xtremesystems.org/forums...-6-8GHz-with-all-cores-active-on-MSI-Titanium

6.8 with all cores active. Personally I do not think it's a waste of time and resources . . . but if it's not your bag then so be it.

More legit... and is near the AMD FX world Record (all cores, 7 Ghz)

 

[SlowSpyder]

i'm more impressed by the average clocks these get on air, thank you. 5ghz with barely over base voltage.

I can do that. :awe:

What a totally pointless exercise. If you can't run the whole chip, then I'd throw out the results.

AMD is still cleaning their clock on these insane overclocks -- An FX-8370 was running all 8 cores at 8.722 Ghz.

http://wccftech.com/stlit-hits-world-record-amd-fx-8370-clocked-87ghz/

I made it to Windows with very limited stability under water with just one module at 5.6GHz. All eight cores at over 8.7GHz, wowzer.

 

[MiddleOfTheRoad]

In terms of raw clockspeed, nothing trumps Vishera. In terms of raw performance, Toppc's 6.8 GHz OC (all cores + HT) on a 6700k would blow away the 8.722 GHz 8370.

I'm not convinced the IPC advantage would overcome the 2 Ghz disadvantage in clock speed under Linux anyways. I know I mention this quite a bit, but the older FX-8350 runs toe to toe with an i7 3770K under Ubuntu -- and in a few benchmarks the FX is actually faster.

BTW, Passmark gives the i7 6700 a score of 10807 -- and an FX 8370 clocks in at 9017 (at stock clocks).
My instincts tell me that extra 2 Ghz overclocking advantage would be enough for the FX to be the faster multithreaded CPU.

 

[MiddleOfTheRoad]

More legit... and is near the AMD FX world Record (all cores, 7 Ghz)

The world record is all 8 cores running at 8.722 Ghz.

 

[MongGrel]

Liquid Nitrogen has more or less always been like CPU drag racing of course.

Short fast runs to get a benchmark.

 

[MiddleOfTheRoad]

I think the fact that it runs on liquid nitrogen is the part that makes it pointless.

Why are AMD chips better at overclocking? I know many of them run at higher stock speeds and probably have lower instructions per clock, but why are they able to clock higher?

The AMD FX was designed with the Pentium 4 mindset -- really long pipelines for extremely high clock speeds. I wouldn't be shocked if an FX is the first to break the 10 Ghz barrier.... They are only 1.278 Ghz away.

 

[SlowSpyder]

The AMD FX was designed with the Pentium 4 mindset -- really long pipelines for extremely high clock speeds. I wouldn't be shocked if an FX is the first to break the 10 Ghz barrier.... They are only 1.278 Ghz away.

http://www.techpowerup.com/204656/the-stilt-drives-amd-fx-8370-to-8722-78-mhz.html

He used liquid nitrogen to get that clock, too. 10GHz seems like a lot, but maybe with great silicon, modules disabled, and even colder liquid helium who knows..? I think these types of things are meaningless other than hitting a number, but still interesting none the less.

 

[Dresdenboy]

These overclocks are so utterly pointless.

What a waste of time and resources.

This could be said about many things in this world. But there are more results besides the fun or competition, like learning to solve technical problems, understanding the technology to make advantage of it, etc.

 

[Blue_Max]

What a totally pointless exercise. If you can't run the whole chip, then I'd throw out the results.

AMD is still cleaning their clock on these insane overclocks -- An FX-8370 was running all 8 cores at 8.722 Ghz.

http://wccftech.com/stlit-hits-world-record-amd-fx-8370-clocked-87ghz/

The energy consumption for the chip and cooling must be astronomical!

...still... impressive!

 

[DrMrLordX]

I'm not convinced the IPC advantage would overcome the 2 Ghz disadvantage in clock speed under Linux anyways. I know I mention this quite a bit, but the older FX-8350 runs toe to toe with an i7 3770K under Ubuntu -- and in a few benchmarks the FX is actually faster.

BTW, Passmark gives the i7 6700 a score of 10807 -- and an FX 8370 clocks in at 9017 (at stock clocks).
My instincts tell me that extra 2 Ghz overclocking advantage would be enough for the FX to be the faster multithreaded CPU.

Maybe, maybe not. 2 GHz in clockspeed isn't that big of a deal when it's only a 28.2% clockspeed advantage. There are plenty of benchmarks where good ol' Haswell has an IPC advantage at or above that in good old Windows. Skylake just ups the ante. In Linux, you're seeing a lot of GCC-compiled stuff instead of MSVC or ICC so . . . take that however you wish.

Also bear in mind that both Vishera and Skylake probably have to make significant memory subsystem compromises to reach core clockspeeds of 8.722 and 6.8 GHz respectively. I'm pretty sure the Skylake machine Toppc worked on to reach 6.8 had to run single-channel DDR4-4200 or so. No idea what memory settings Stilt had to use. Regardless, proper clockspeed scaling takes a big hit when clockspeeds get out ahead of the memory subsystem. At that point it's a contest to see whose cache subsystem can best cope with inadequate memory bandwidth. Skylake will gain over Vishera under those conditions.

So I would expect Skylake to win most benchmarks in any kind of comprehensive benchmark suite run under those settings, assuming either machine had the ability to stay clocked so high for any extened period of time (which is probably untrue for either machine).

 

[MiddleOfTheRoad]

So I would expect Skylake to win most benchmarks in any kind of comprehensive benchmark suite run under those settings, assuming either machine had the ability to stay clocked so high for any extened period of time (which is probably untrue for either machine).

And I wouldn't. Ivy and Haswell's major performance advantage over AMD is at single threaded apps. Ivy generally trails the FX at multithreaded. Look at these benches -- then add an additional 4.1 Ghz of clockspeed to the FX (versus the 4.6 Ghz FX 8350 in these benches). I doubt that even the hottest Skylake chips can touch those types of multithreaded performance numbers.

Granted, the FX is pretty lame at single threaded stuff. But -- overclocking the thing to close to 9 GHz.... The multithreaded stuff has got to be insane since performance scales well on these chips and the sheer brute force of running all 8 cores at more than twice stock clock speed. Skylake probably is better at single threaded, but I seriously doubt it can touch a nearly 9 Ghz Vishera at multithreaded stuff. I think many people underestimate the actual capabilities of FX chips (because they are usually crippled by the Intel compiler under Windows). FX chips are multithreaded monsters under Linux. I suspect DirectX 12 will change a lot of people's minds about Radeons, as well.

 

[cmdrdredd]

well then you tell me. six reviews out of six on 4.8~5Ghz with 1.30 on youtube. did they all get cherry picked samples?

None of the review sites I read could get that with a retail chip.

 

[Sweepr]

Those are outdated results, in the latest benchmarks a higher clocked Vishera FX8370 can't keep up with 4C/4T Core i5 6600K.

Now answering the question, no doubt 6.8GHz Core i7 6700K would be faster, ST or MT.

 

[MiddleOfTheRoad]

Those are outdated results, in the latest benchmarks a higher clocked Vishera FX8370 can't keep up with 4C/4T Core i5 6600K.

Now answering the question, no doubt 6.8GHz Core i7 6700K would be faster, ST or MT.

First, you just completely moved the goalposts. Only 1 of those benches is the same as I posted. Nice way of Cherrypicking completely different benches. FX still looks to be faster on the benches I was using. We are talking a 120% overclock on the FX versus a 70% overclock on the Skylake.

 

[Sweepr]

Plenty of MT benchmarks for you to browse (from the same source of your old results) so you can't say I'm cherry-picking: http://www.phoronix.com/scan.php?page=article&item=intel-6600k-linux&num=1

I'll save you some time: Vishera was slower than 4C/4T Core i5 6600K in all of them. Time to update your ad nauseam 'Vishera matches 4C/8T Intel @ Linux MT claim'.

Also if you move the goalpost to Windows, just in case: According to Hardware.fr's overall results: Core i7 6700K (4.0GHz Turbo @ MT loads) is 18% faster than FX9590 (~20% higher clocks, 4.7-5.0GHz). 28% clockspeed advantage would still put Vishera behind Skylake @ MT while being demolished in ST.

 

[DrMrLordX]

FX chips are multithreaded monsters under Linux.

I think you're seeing things with rose-colored glasses. That FX isn't going to scale perfectly when overclocked to 8.722 GHz, nor is the Skylake going to scale perfectly all the way up to 6.8 GHz. What scaling does happen will be heavily-influenced by cache since the memory speeds/timings will be all screwed up on both platforms. I'll let you argue with Sweepr about benchmarks at lower clockspeeds. My point is, you can't just take existing scores from "normal" clockspeed ranges and extrapolate useful information about how either chip will perform when pushed well beyond all reasonable clockspeed limits.

It is questionable as to whether either system was stable enough to run a full suite of benchmarks before wiping out all the LN2 they had on-hand. What benchmarks they would have run would have been Windows benchmarks anyway. MS operating systems were in use for both suicide runs.

I would like to see the two systems run some Linux software, since I do use Linux, but I wouldn't hold my breath waiting for it.

 

[MiddleOfTheRoad]

Plenty of MT benchmarks for you to browse (from the same source of your old results) so you can't say I'm cherry-picking: http://www.phoronix.com/scan.php?page=article&item=intel-6600k-linux&num=1

I'll save you some time: Vishera was slower than 4C/4T Core i5 6600K in all of them. Time to update your ad nauseam 'Vishera matches 4C/8T Intel @ Linux MT claim'.

You're still glossing over the fact that one of these chips maxed out at a 70% overclock while the other managed a 120% boost in clockspeed. Even with diminishing returns as you increase clock speeds -- the original benches would indicate the FX will scale very far.

This reminds me when people said that an FX 8320 wasn't going to post i7 numbers daily on World Community Grid.... Well, my 4 million points proves otherwise. I'm telling you right now -- the Intel 6600K doesn't have a prayer IMO against an FX-8370 overclocked to 8.722 Ghz doing integer calculations. Just ain't happening -- the nearly 9 Ghz Vishera will whip its ass. It's much the same reaction to single threaded -- you'd probably need to overclock a Vishera to 15 Ghz (I'm exaggerating, but still) to match the 6600k's single threaded performance....

 

[Sweepr]

You're still glossing over the fact that one of these chips maxed out at a 70% overclock while the other managed a 120% boost in clockspeed. Even with diminishing returns with the increased clock speeds -- the original benches would indicate the FX would be faster at multithreaded.

Just like people said that an FX 8320 wasn't going to post i7 numbers daily on World Community Grid.... Well, my 4 million points proves otherwise.

Lowly 4C/4T Skylake beating Vishera @ Phoronix Linux suite (i7 would 'whip its ass') and Core i7 6700K faster than 20% higher clocked FX9590 (Hardware.fr) would like to have a word with you. And as DrMrLordX correctly pointed scaling wouldn't be perfect for either, so it's a pointless discussion.

the Intel 6600K doesn't have a prayer against an FX-8370 overclocked to 8.722 Ghz doing integer calculations. Just ain't happening -- the nearly 9 Ghz Vishera will whip its ass.

Well color me suprised, here was I thinking we were comparing 6.8GHz Core i7 6700K to the 8.7GHz FX. Or maybe you just moved the goalpost again.