4GHz i7 920

[Idontcare]

4GHz i7 920 3DMark06 Screenshot

4GHz i7 920 SPi 1M Screenshot

4GHz i7 920 Everest Screenshot

link to source forum thread

 

[disports]

damn, sick.

 

[Cookie Monster]

Impressive but i doubt retail will OC like that. Thats an ES chip, i.e probably with an unlocked multipler making OCing easier.

 

[myocardia]

Too bad the only reason it scores ~7,400 in the 3dMark06 CPU score is because it's able to run 8 threads. It's going to be a monster for things like protein folding, though, and probably most other DC projects. BTW, I'm utterly surprised that they haven't "leaked" any benchmarks with the only game ever sold that actually benefits from >4 cores-- M$'s FSX.

 

[Idontcare]

Originally posted by: Cookie Monster
Impressive but i doubt retail will OC like that. Thats an ES chip, i.e probably with an unlocked multipler making OCing easier.

True it could be a cherry-picked sample, but unless Intel changed their ES procedure then the multiplier is not unlocked like the extreme chips.

You can only lower the multiplier on ES chips, same as you can do on non-extreme chips.

I just thought it was a good sign. If ES 920's were only overclocking to 3.2GHz it would mean something different versus overclocking to 4GHz.

But this also means something about the reference clock, unless the mobo itself is cherry-picked as well this means that those 940's should be getting all the further above 4Ghz, at least to whatever the CPU's silicon can handle instead of being limited by the reference clock top speed.

 

[OCGuy]

Originally posted by: Cookie Monster
Impressive but i doubt retail will OC like that. Thats an ES chip, i.e probably with an unlocked multipler making OCing easier.

An unlocked multiplier on a C2D was a joke. Who needs one when you can just up the FSB? I know FSB is a thing of the past when it comes to i7, but I dont think the multi is going to be needed in order to OC.

 

[AuDioFreaK39]

Originally posted by: Ocguy31

Originally posted by: Cookie Monster
Impressive but i doubt retail will OC like that. Thats an ES chip, i.e probably with an unlocked multipler making OCing easier.

An unlocked multiplier on a C2D was a joke. Who needs one when you can just up the FSB? I know FSB is a thing of the past when it comes to i7, but I dont think the multi is going to be needed in order to OC.

On the contrary, I've only seen QPI increases of less than 30MHz for 4.5GHz overclocks. The multi will still become a critical factor in overclocking Nehalem chips - at least the Extreme Editions.

If the i7 920 runs at 133MHz x 20, then upping the QPI to 200MHz will give ~4.01GHz. The key is recognizing the 133MHz default clock speed.

 

[Extelleron]

If a Bloomfield takes 1.472V just to be stable at 4GHz in 3D Mark, then I can't say I am impressed.

Nehalem is looking good for those who run their CPUs @ stock, but for those who overclock the leaked results don't look so hot.

Maybe it will be better with a newer revision or perhaps the CPU tested is not the launch revision (although it is the newest we have seen.) IMO at least it will be best to wait until 2009 to buy a Nehalem platform unless you don't OC or you just need the performance now.

 

[BTRY B 529th FA BN]

Overclocking this will be interesting. I was never 100% for sure on OCing my AMD systems so this will a back to school adventure

 

[covert24]

the super PI isnt really that amazing. as my E8400 can do it in 11s....

 

[n7]

Originally posted by: Ocguy31

Originally posted by: Cookie Monster
Impressive but i doubt retail will OC like that. Thats an ES chip, i.e probably with an unlocked multipler making OCing easier.

An unlocked multiplier on a C2D was a joke. Who needs one when you can just up the FSB? I know FSB is a thing of the past when it comes to i7, but I dont think the multi is going to be needed in order to OC.

I take it you've never played with quads?

Multi is a very big deal for quads, & i'd expect it to continue to be.

Considering we're looking at an entirely new platform, i don't expect OCing to be a piece of cake with the initial motherboards either, but i might be wrong (i hope i am )

Originally posted by: Extelleron
If a Bloomfield takes 1.472V just to be stable at 4GHz in 3D Mark, then I can't say I am impressed.

Nehalem is looking good for those who run their CPUs @ stock, but for those who overclock the leaked results don't look so hot.

Maybe it will be better with a newer revision or perhaps the CPU tested is not the launch revision (although it is the newest we have seen.) IMO at least it will be best to wait until 2009 to buy a Nehalem platform unless you don't OC or you just need the performance now.

I would somewhat agree...assuming that vcore level is being read correctly.

Originally posted by: covert24
the super PI isnt really that amazing. as my E8400 can do it in 11s....

:laugh:

There's a lot less L2 cache on these chips, which is one thing SuperPi loves.

I'm amazed it did that well in that useless bench actually.

Look at the Everest results.

Those are some sick numbers for bandwidth & latency :Q
Absolutely fantastic.

 

[Idontcare]

Originally posted by: n7
Look at the Everest results.

Those are some sick numbers for bandwidth & latency :Q
Absolutely fantastic.

They are pretty sick, aren't they! <4ns L3$ latency, <31ns Ram latency. The mind boggles. And that ram bandwidth, >15GB/s! Phenomenal.

 

[covert24]

yea i realize super PI is worthless. But now that i look at that everest benchmark its out of this world. Good stuff by intel

 

[VirtualLarry]

The all-important question: what mobo are they using? I guess this is also confirmation that Nehalem overclocks.

Edit: Why did he blur out the CPU bus speed and multi? Doesn't want anyone to know it's an unlocked chip or something, or is it unlocked?

 

[Idontcare]

Originally posted by: VirtualLarry
The all-important question: what mobo are they using? I guess this is also confirmation that Nehalem overclocks.

Edit: Why did he blur out the CPU bus speed and multi? Doesn't want anyone to know it's an unlocked chip or something, or is it unlocked?

Most of these "leaks" are sanctioned by seeded reviewers who are given permission to "leak" screenies on various forums provided there is a drip-drip-drip to the data.?

No full disclosure of everything so the next guy sanctioned to drum up some more excitement over a screenie next week can have a chance.

Your ability to restrain yourself with the hardware and NDA in this cycle determines your likelihood of being seeded as a sanctioned leaker for the next hardware cycle (westmere, etc). So the folks getting to leak stuff this cycle must have been good to their NDA word on their last cycle with Penryn.

At any rate sure its viral and we aren't supposed to like that because of its manipulative nature, but if you want to glean anything from the web about Nehalem (such as "will 920's OC?) in advance of official release then you have to be on the lookout for these sanctioned leakers rebelling against the establishment and putting up the screenies "despite the NDA"...power to the viral marketeers I guess, but at least we get some drip-drip-drip info nonetheless.

? no, I can offer zero proof that this is the way the world works, feel free to disagree with me and reject my version of reality

 

[OCGuy]

Originally posted by: n7

Originally posted by: Ocguy31

Originally posted by: Cookie Monster
Impressive but i doubt retail will OC like that. Thats an ES chip, i.e probably with an unlocked multipler making OCing easier.

An unlocked multiplier on a C2D was a joke. Who needs one when you can just up the FSB? I know FSB is a thing of the past when it comes to i7, but I dont think the multi is going to be needed in order to OC.

I take it you've never played with quads?

Multi is a very big deal for quads, & i'd expect it to continue to be.

Considering we're looking at an entirely new platform, i don't expect OCing to be a piece of cake with the initial motherboards either, but i might be wrong (i hope i am )

Ive had a Q9450...lets take a look at quads.

QX9770 - $1499. 400*10 = 4ghz
Q9650 - $500 - 450 * 9 = 4ghz

For most people, the unlocked multi on Penryns is a gimic......

 

[Lonyo]

Call me crazy, but isn't Nehalem going to be absolute perfection for Photoshop and its ilk?
Imagine 24GB RAM (6x4GB) or even 12GB (6x2) with say 8+GB made into a RAMdisk with 15GB/s. Plus lots of cores.
4 cores, 8 threads, 6 RAM slots = infinite possibilities.

 

[Idontcare]

Originally posted by: Ocguy31
QX9770 - $1499. 400*10 = 4ghz
Q9650 - $500 - 450 * 9 = 4ghz

For most people, the unlocked multi on Penryns is a gimic......

I agree. The only reason I would buy an extreme edition for the unlocked multiplier is if I were using something other than air cooling. Like a vaporphase rig which would take that QX9770 to 5GHz with the higher multiplier whereas the Q9650 probably isn't going much higher than 4GHz on air or on vaporphase cooling.

 

[AuDioFreaK39]

Originally posted by: Lonyo
Call me crazy, but isn't Nehalem going to be absolute perfection for Photoshop and its ilk?
Imagine 24GB RAM (6x4GB) or even 12GB (6x2) with say 8+GB made into a RAMdisk with 15GB/s. Plus lots of cores.
4 cores, 8 threads, 6 RAM slots = infinite possibilities.

too bad Photoshop CS4 just switched over to GPU rendering

can't wait to run it on my GTX 280

 

[rge]

How do you get exactly 4010 mhz from integers 133 or higher multiplied by integers 20 to 30 (or even with half multi's). I cant get there in excel. Also how did he get 267 stock, is that 133.3?x20, 266 would be 133x20. I can get 4011 using 21x191 (if 133) or 4010.6 using 22x182.3 (if 133.3), but he is showing 4010. Cant seem to get any closer, especially not using higher multipliers.

Edit: Found it, if it is actually 133.4, which explains the 2.67, and 25x160.4 is only way to get exactly 4010, and makes sense with those claiming only 30mhz increase so far on the QPI.
http://diy.pconline.com.cn/cpu...s/0809/1426985_10.html

 

[KingstonU]

How much $$$ for this rig that got it to 4GHz? Including mobo and RAM (3 sticks of DDR3 right?) and everything

 

[secretanchitman]

we dont have a clue about mobos yet (im assuming at least $200 cause it'll be a brand new mobo chipset), and at least $300 for the low end bloomfield cpus, and for ram, it fluctuates day to day...

 

[Idontcare]

Originally posted by: rge
Edit: Found it, if it is actually 133.4, which explains the 2.67, and 25x160.4 is only way to get exactly 4010, and makes sense with those claiming only 30mhz increase so far on the QPI.
http://diy.pconline.com.cn/cpu...s/0809/1426985_10.html

It's really 133.33333333333333333333...but the way you end up with software programs reporting numbers that seem rounded in random ways is that the chipset is usually using spread-spectrum which intentionally shifts the frequency around a bit for EMI reasons.

Most BIOS's offer the option of disabling spread-spectrum. This particular person either elected to leave it enabled (and it shifted to 133.4) or was not presented with the option of disabling it in their BIOS.

 

[Tempered81]

Originally posted by: KingstonU
How much $$$ for this rig that got it to 4GHz? Including mobo and RAM (3 sticks of DDR3 right?) and everything

> $800

 

[AuDioFreaK39]

Originally posted by: rge
How do you get exactly 4010 mhz from integers 133 or higher multiplied by integers 20 to 30 (or even with half multi's). I cant get there in excel. Also how did he get 267 stock, is that 133.3?x20, 266 would be 133x20. I can get 4011 using 21x191 (if 133) or 4010.6 using 22x182.3 (if 133.3), but he is showing 4010. Cant seem to get any closer, especially not using higher multipliers.

Edit: Found it, if it is actually 133.4, which explains the 2.67, and 25x160.4 is only way to get exactly 4010, and makes sense with those claiming only 30mhz increase so far on the QPI.
http://diy.pconline.com.cn/cpu...s/0809/1426985_10.html

omg finally some fully leaked specs! thanks for the find man, and thanks for taking the time to calculate the QPI bus speed with precision in order to confirm the findings.