5Ghz air done Core i7 Extreme

[[XC] riptide]

What does 'x = -2' stand for? The very bad overclocking chips or the chips that just don't work. Is the starting point the standard clock frequency of your chip or is it the lowest stable frequency of a certain batch of processors.

Never mind the legend on the scales. Its a standard Gausian curve. You can apply it to population of lots of things including humans in terms of IQ, height etc.

If we applied it to CPU's, the centre of that graph where it is highest represents the average CPU. standard overclock for any given chip. This would be where most CPU's fall into. AS you get closer to the cherry picked chips, you obviously have less numbers of them... and you move right on the graph. The graph falling represents less numbers of them. I expect that if you were to mark on that graph, this particular chip that Francois and charles played with, it would be on the extreme right.

The dodgy (E8500 4.3Ghz at 1.49vcore? GTFO! ) E8500 that i have for example would sit somewhere on the left of centre.

[Marvin_The_Martian]

One of these?


Sorry couldn't resist, having a very bad day and needed an ego boost

[[XC] riptide]

I think its already been established that these calibre chips will appear in small numbers naturally, as have the Golden Intels from times gone by. Francois was just in a position to get his dirty hands on a few of them.

Just because most of us can't get these chips as easily as they have been got for this 'demo', does not mean

1) That they won't appear in retail

or 2) invalidates the clocks achieved.

Srsly... the chips have been brought outside of Intels corporate offices into the Enthusiast realm, in the form of FUGGER, and displayed, as opposed to a Manufacturer supervised and run demo.

We should be thankful that we have an Intel employee (albeit an arrogant one ) who can show us such things and is 'on our side'.

[NBF]

Im wondering how those golden chips are binned.

Do you test then with a fixed Vcore and look for those with the higgest current draw? Trying to find the "leaky" ones?

If thats the way to do it, maybe they could be selected just measuring resistance between some of the 1366 pins, making it quicker and easier to find "the neddle in the haystack".


PS. Im just thinking out loud, maybe very wrong about it. Also dont expect you guys can answer my question

[Praz]

I guess something is not being written as intended. The following quotes from Drwho? do not read the same as the posts that would have us believe the chip was a cherry-picked retail sample.

well, i said it at the beginning, those are special chips. I started with this.
I was very very very very clear from the beginning. Those are special chips, and I don t want to give more details. They are using the today's selling chip settings.

Special chips programmed with parameters from chips we can buy today. If these were retail chips would not these parameters be the correct ones already?

wait wait wait ... you can always find a side effect for everything ... The demo started with the idea that Manufacturer demos are useless, you just made my point !!!! thanks!

The above quote was made in response to a post stating that chips in the retail market would not be able to match this performance.

Those are CPUs I have in my lab to play with, The point is clear , Manufacturer can do demos that does not reflect always what gets to the market! This is EXACTLY the point I wanted to make from the beginning.

I WARNED many time.

This demonstration made a point that what a manufacturer can showcase may not be true in retail.

When I use special chips, I say it loud and clear, when it is normal chips, you ll see no special notes. I oppose very hard the special chip demos inside my own company, but I had to make a point that it is easy for a manufacturer to put a demo that shine, without really reflecting the market.

Once again referencing that this demo does not reflect real word retail results.

If I read Dr. Who?'s comments correctly then yes, these are chips you 'could" see in retail. That they are cherry picked but that normally these chips would have gone "down the line" been boxed and sold..

This was not a canned demo, it was a hard push to resolve issues and many things pointed out in this thread were covered to make processors even better.

So what's the deal with the chip? Is it as Drwho? implies a specimen that is capable of performance exceeding the chips in the current retail market? Or as Movieman and FUGGER suggest a super chip pulled from the retail line? Shouldn't be difficult to state one way or the other. I think that is all that is being asked in this thread.

[Gautam]

Temporarily locking, this is more of a mess than I can personally tolerate.

[Charles Wirth]

Thanks Gautam, my apologies everyone else. The rivalry is alive and kicking

Best binned chips would go into 965 cores, leakage to some point cores are screened out into the next bin down.

With Francois looking at what we do and knowing what we need now, that combined with an improved process later on will make for an interesting chip.

I need to chase down Duck, he has a good chip.

[Vapor]

FWIW....the reason some people's questions are not being answered, especially about chip specifics, is because those who know the details can't elaborate on the details.

They're not trying to pull a smoke and mirrors thing here or presenting any sort of "argument from silence" it's just that they're not allowed to talk specifics.

[NapalmV5]

great clocks regardless

would love to see what 5ghz or max do @

nuclearus
http://www.xtremesystems.org/forums/...d.php?t=202576

fritz chess
http://www.xtremesystems.org/forums/...ighlight=chess

[Sampsa]

I see some posts were removed.. maybe better that way

[sdumper]

Cherry chip? Was anything done to the chip?
I remember Frances saying in an AMD thread that he was going to have a chip modded with some features turned off so that he could post some high numbers.

-edit- never mind I just read the first post and didnt read the last few...looks like other people have asked the same questions.

Good to see AMD and Intel battle it out again :-)

[vmsein]

Nothing wrong with demonstrating what will one day be possible for everyone. Nicely done, fellas!

[vab206]

so, from what im hearing (judging by a vid on the nehalem by intel CEO) he says that 8 core configs will ship in 2008. But, the question i have is when will the number of cores come to an end. in my opinion 8 cores seems to be a bit much, 16 simultaneous threads is a bit insane. what kind of software/OS out there can utilize something like that to its fullest extent to start with? Why dosent intel make a nehalem 2 or 4 core just for gaming? 2 cores seems to be all that most games utilize anyways and with nehalem's architecture it can process 4 threads, so why not drop the on die cores to 2 and raise clock speeds?

[verndewd]

well thats just it in the end the desktop os will be alot more like server os's in that it can sustain a larger grouping of cores. 4x4 is what? and were talking about current maintsream server configs. higher core counts do an immense amout of good for servers and in the end will enable apps to take singular advantage of cores thereby freeing resources and speeding multiple processes up. alot of cores means alot of freedom and alot of speed.

sun has even more cores on its server cpu's I think its already at 16 and heading up not down. lets say you pack 1k cores on a die in a desktop, there isnt a whole lot that cant be done on that hypothetical system.

[Drwho?]

well thats just it in the end the desktop os will be alot more like server os's in that it can sustain a larger grouping of cores. 4x4 is what? and were talking about current maintsream server configs. higher core counts do an immense amout of good for servers and in the end will enable apps to take singular advantage of cores thereby freeing resources and speeding multiple processes up. alot of cores means alot of freedom and alot of speed.

sun has even more cores on its server cpu's I think its already at 16 and heading up not down. lets say you pack 1k cores on a die in a desktop, there isnt a whole lot that cant be done on that hypothetical system.

http://www.tomshardware.com/news/win...-gpu,6645.html
Homogeneous computing is coming, it is just a matter of few years

[Marios]

LOL Is Intel about to give us a new 975 Extreme Edition stepping 5?
Then those posts about Phenom II overclockability are not all about cherry picked and tweaked CPUs.
Another explanation is that a programmed chip to a newer edition circumvents the TLB BIOS restriction.
Or what else?
975 is obvious in the video anyway.

[BenchZowner]

LOL Is Intel about to give us a new 975 Extreme Edition stepping 5?

It doesn't take a lot of thinking or wisdom for anyone who has been "watching" the computer market the last two years that Intel is releasing a "refresh" of their flagship processor some months after the release of their initial flagship model.
You remember the QX9650 & QX9770 deal, and the QX6700 & QX6800.
It's surely going to happen again.
Without having any particular info from inside, I'd say that we're going to see a new Extreme Edition Core i7 with a new revision of the core ( C1 ) in late January, Mid February.

Another explanation is that a programmed chip to a newer edition circumvents the TLB BIOS restriction.

The TLB fix is being applied already through BIOS.

To get rid of the errata physically you need to physically change the design of a part of the CPU.

[vab206]

well thats just it in the end the desktop os will be alot more like server os's in that it can sustain a larger grouping of cores. 4x4 is what? and were talking about current maintsream server configs. higher core counts do an immense amout of good for servers and in the end will enable apps to take singular advantage of cores thereby freeing resources and speeding multiple processes up. alot of cores means alot of freedom and alot of speed.

sun has even more cores on its server cpu's I think its already at 16 and heading up not down. lets say you pack 1k cores on a die in a desktop, there isnt a whole lot that cant be done on that hypothetical system.

The end... lol, there will be no end. The amount of cores will continue to grow while gamers still buy only what they need to play their games. im not talking about servers at all. ill believe W7 is a better gaming platform then XP when i see it.

what im simply trying to say is that wouldnt a dual core machine that could clock higher be more beneficial to gamers then a 4core 8 thread machine or a 8 core 16 thread machine? gamers use 2 MAYBE 3 programs at once. those are the game their playing, ventrillo/teamspeak, and occasionally a browser to look something up. lets face it, a processor like the nehalem is not needed to compete these tasks with ease.

if intel would make a dual core nehalem im nearly 100% positive that it would be a strong seller amongst the growing gaming community. lets be honest, if all you did was play a game when you were on the computer wouldn't you buy a good core2 over a nehalem?

[CryptiK]

Great results, good work Fugger & Francois, that must have been a fun weekend.

I don't understand why there are so many people complaining that CPU's of this caliber may not be available commonly in retail. Not many people even push CPU's that hard, it's as if those upset by this feel like they have been prevented from attaining something that they neither have the skill nor inclination to do anyway. There are comparatively few people here who really push hardware to the limits either on air, water or extreme cooling. Those that do have the ability, finances and contacts that enable them to find and use the best CPU's available anyway. Some people already have contacts at Intel which enable them to get cherry picked ES CPU's - this has been happening forever, it's nothing new at all.

World records have always been set with specially picked CPU's capable of clocks above what most can achieve, this is just taking that to a slightly higher level, as Intel themselves picked the chips in question. If these CPU's were not going to be 920's/940's or 965's, but perhaps an as yet unreleased version (eg 975's) there is nothing wrong with that, in time CPU's of higher spec will most likely be released if yields are sufficient, and people can try and find the best ones by the same means they have already employed to find the best 775's and already released i7's.

This was demo to show the theoretical maximum achievable from this architecture, using CPU's selected by employing a binning process no one but intel has the ability to do, and as such is not a reflection of what most people can expect from retail samples. This is the nature of a manufacturer demo. Francois made this clear.

[[XC] riptide]

Great results, good work Fugger & Francois, that must have been a fun weekend.

I don't understand why there are so many people complaining that CPU's of this caliber may not be available commonly in retail. Not many people even push CPU's that hard, it's as if those upset by this feel like they have been prevented from attaining something that they neither have the skill nor inclination to do anyway. There are comparatively few people here who really push hardware to the limits either on air, water or extreme cooling. Those that do have the ability, finances and contacts that enable them to find and use the best CPU's available anyway. Some people already have contacts at Intel which enable them to get cherry picked ES CPU's - this has been happening forever, it's nothing new at all.

World records have always been set with specially picked CPU's capable of clocks above what most can achieve, this is just taking that to a slightly higher level, as Intel themselves picked the chips in question. If these CPU's were not going to be 920's/940's or 965's, but perhaps an as yet unreleased version (eg 975's) there is nothing wrong with that, in time CPU's of higher spec will most likely be released if yields are sufficient, and people can try and find the best ones by the same means they have already employed to find the best 775's and already released i7's.

This was demo to show the theoretical maximum achievable from this architecture, using CPU's selected by employing a binning process no one but intel has the ability to do, and as such is not a reflection of what most people can expect from retail samples. This is the nature of a manufacturer demo. Francois made this clear.

I wish everyone would understand this.

[Marios]

......

This was demo to show the theoretical maximum achievable from this architecture, using CPU's selected by employing a binning process no one but intel has the ability to do, and as such is not a reflection of what most people can expect from retail samples. This is the nature of a manufacturer demo. Francois made this clear.

So you think we saw the max from the new CPU?
Wait and see.
Besides showing the world the capabilities of Core i7, I think Francois wanted to take the advice of his friend - an extreme overclocker - "over current known issues" in overclocking.
This benefits the rest of us and I like Intel does it this way.
They produce better products for everyone.

[Charles Wirth]

What we see is similar to what auto manufactures did many years ago, so many similarities between the two.

People need to understand that most of the top guys have sponsors. Much like racing, sponsors will pick who is active and has good communication, most of all deliver good results.

I have been showing chips for years, its not unlike me so it should be no surprise there. I see overclocking getting bigger in 2009 and first up is CES.

[mXpS]

What we see is similar to what auto manufactures did many years ago, so many similarities between the two.

People need to understand that most of the top guys have sponsors. Much like racing, sponsors will pick who is active and has good communication, most of all deliver good results.

I have been showing chips for years, its not unlike me so it should be no surprise there. I see overclocking getting bigger in 2009 and first up is CES.

I'm hoping you guys have that Ace up your sleeve and you plan on taking it out at CES

[Charles Wirth]

4.6Ghz wprime on air with stock i7 820 heatink done, playing more on air.

The Coolermaster heatsink is kinda sucky, has horrible contact if done right and no contact if you are not paying attention. It doesnt fit correctly with the retrofit 1366 brackets.

[hipro5]

This seems to be a blast IF most of the "new ones" could - at least do - the frequencies of the QX9650/9770 on air......