Saturday, October 11th 2008

Core i7 965 XE Races Through to 4.20 GHz

It's high time to break the dogma. Core i7 965 Extreme, Intel's next generation flagship desktop processor based on the Nehalem architecture, does indeed overclock satisfactorily. IT OC Taiwan overclocked the chip, which sports an unlocked FSB multiplier, to a respectable 4.20 GHz, at a FSB speed of 200 MHz, and multiplier value of 21.0xFSB. A vCore setting of 1.72 V was used, which is above the danger-mark, taking Intel's own warnings into account. The CPU was aided by dual-channel DDR3-1600 memory operating at timings of 8-7-7-24. It provided a tested bandwidth of close to 16,000 MB/s. The CPU took 3DMark Vantage for a spin, with 3596.76 plans/s in CPU Test 1, and 32.87 steps/s in Test 2. The chip used in the attempt was an engineering sample.
Source: IT OC Taiwan
Add your own comment

70 Comments on Core i7 965 XE Races Through to 4.20 GHz

#26
farlex85
Tatty_One
I can get 4.2gig on 1.35v :D .....taking into account that Nahlem is of course faster than Wolfdale clock for clock, lets say I can get 4.4gig on 1.425v ;) This kinda worries me abit, mainly cause it suggests to run at say 4gig 24/7 you are still gonna need like 1.6V.....thats just un-realistic.
Even so, you only have half the logical cores and less than half the memory bandwidth. But, then again, your proc was half the price, and those won't matter in 90% of apps anyway (well, maybe 80). :) And yeah, doesn't seem to practical.
Posted on Reply
#27
kid41212003
But, seriously, do we need Quad-cores over 4GHz?
If the slowest i7 core able to OC to 3.2GHz and with voltage under 1.3v, then I think it's already too good to kick AMD ass.
And according to the Vantage CPU score, Nehalem prove to be faster than the current QX ~25% clock vs clock.

We do not need Quad-core 4GHz for gaming, and I don't think there are many people like to run their CPUs 24/7 with a voltage over 1.4v
Posted on Reply
#28
zithe
kid41212003
But, seriously, do we need Quad-cores over 4GHz?
If the slowest i7 core able to OC to 3.2GHz and with voltage under 1.3v, then I think it's already too good to kick AMD ass.
It's too good to be better? What? XD
Posted on Reply
#29
SimFreak47
Bloomfield with 1.7 volts = Boomfield!

Dude, that thing would probably explode with all those volts!!!
Posted on Reply
#30
kid41212003
zithe
It's too good to be better? What? XD
I mean, It's already better than AMD with a huge gap, even If it's not OC good.
Posted on Reply
#31
ShadowFold
kid41212003
I mean, It's already better than AMD with a huge gap, even If it's not OC good.
These things are barely any faster than a QX9770...
Posted on Reply
#32
KBD
not surprised by the results at all. I expected to be around 4 gig. I think the IMC has a lot do with this. What kind of cooling was used, BTW? Air, water?

And this is extreme version with unlocked multi, i wonder how the lower chips will do?
Posted on Reply
#33
kid41212003
ShadowFold
These things are barely any faster than a QX9770...
Check out some 3DMark Vantage CPU scores here:

Link.

I can understand why you think so, but let's me tell you something, which games at this moment can fully support 4 cores and 8 threads?
Posted on Reply
#34
TheGuruStud
kid41212003
Check out some 3DMark Vantage CPU scores here:

Link.

I can understand why you think so, but let's me tell you something, which games at this moment can fully support 4 cores and 8 threads?
Not to mention being synthetic that it's practically all BS anyway.
Posted on Reply
#35
farlex85
TheGuruStud
Not to mention being synthetic that it's practically all BS anyway.
No it's not, synthetic benchmarks are intended to do the same things that would be required of your hardware during normal computing tasks, and even if not it still is a good indicator of raw power, although vantage is probably not the best, especially when considering game performance.

Also I wonder about the stepping, if this is an unfinished stepping the results may be less than at launch.

kid41212003
I mean, It's already better than AMD with a huge gap, even If it's not OC good.
The denab has been show at these speeds, if they have closed the clock for clock gap then those may actually be closer than amd has been in a while (since before core 2), at least w/ apps that don't utilize 8 threads (most of them). And I would wager those come in at much less than $999......
Posted on Reply
#36
aquariuz
Fsb?

sorry guys, im not very well informed about these things, but I heard before dat the nehalems no longer uses FSB but instead something called quickpath. Alrite, but why was FSB mentioned here though? is there something im missing here? :banghead:

thnks guys.
Posted on Reply
#37
karlotta
I think somebody missed the real score. Look at the Everest cache and Memory bench...
No Wolfy is getting 16000, and 26.4 Lat. If you cant use that then stick with your Benchmark Game and clock till you drop. The X48 and p45 will be around for awhile ,though they are getting stale.
Posted on Reply
#38
farlex85
aquariuz
sorry guys, im not very well informed about these things, but I heard before dat the nehalems no longer uses FSB but instead something called quickpath. Alrite, but why was FSB mentioned here though? is there something im missing here? :banghead:

thnks guys.
Cpu-z has just not been updated most likely. They may still call it fsb, even though the architecture is completely different, but most likely after release it will be changed to qpi.

karlotta
I think somebody missed the real score. Look at the Everest cache and Memory bench...
No Wolfy is getting 16000, and 26.4 Lat. If you cant use that then stick with your Benchmark Game and clock till you drop. The X48 and p45 will be around for awhile ,though they are getting stale.
That is of course the biggest improvement nehalem brings to the table, improved bandwidth and latency, truly utilizing ddr3 (actually what's really surprising there is the ddr3 is only running in dual channel). The performance of that system though, when matched w/ what it will cost ($1k for the cpu, $500 for the mem, untold for the board, untold for the cooling) is not good at all though. Still, promising for the future, but amd has some pretty good bandwidth themselves....
Posted on Reply
#39
mmaakk
I didn't see mentions about the TEMP.

With 1.7Vcore I guess this baby is running HOT HOT!
Posted on Reply
#40
Hitsugaya_Toushirou
aquariuz
sorry guys, im not very well informed about these things, but I heard before dat the nehalems no longer uses FSB but instead something called quickpath. Alrite, but why was FSB mentioned here though? is there something im missing here? :banghead:

thnks guys.
Intel's QuickPath Interconnect is still considered a Bus since the definition of a 3rd generation bus includes AMD's HyperTransport, which is more or less equivalent to Intel's Quickpath Interconnect. And since it is the bus which connects the CPU to the Northbridge, it is still called the FSB although a different kind of FSB to the traditional old architecture.
AMD has had their HyperTransport since 2001 but it has always been called a FSB so it is the same thing here.
Posted on Reply
#41
btarunr
Editor & Senior Moderator
aquariuz
sorry guys, im not very well informed about these things, but I heard before dat the nehalems no longer uses FSB but instead something called quickpath. Alrite, but why was FSB mentioned here though? is there something im missing here? :banghead:

thnks guys.
Bus speed x FSB Multiplier = CPU Clock Speed

Bus speed x CPU-NB interconnect multiplier = Interconnect speed

Example: Let's take Phenom X4 9750

200 MHz x 12.0 = 2400 MHz (clock speed)

200 MHz x 9.0 = 1800 MHz (HyperTransport link speed (equals 3600 MT/s))

Now, your altering the bus speed, affects both the clock-speed and interconnect speed. End of the day, the interconnect (HyperTransport for AMD K8 onwards, QuickPath for Intel Nehalem onwards) handles data transfers between the CPU and the northbridge.
Posted on Reply
#42
PVTCaboose1337
Graphical Hacker
1. I cannot afford.
2. WTF 1.72 volts, kill me now.
3. Multi surprised me. Alot. 21x wow.

Anyway, interesting offering to say the least. I think I might wanna do WC'ing if I go this route.
Posted on Reply
#43
eidairaman1
The Exiled Airman
and there are naysayers of deneb requiring 1.6 is too much, pfffttt, this just proves those people wrong, especially when intel specified anything higher than 1.6 for their CPUs is too much and will kill them quick.
Posted on Reply
#44
Guitar
These overclock very nicely.
Posted on Reply
#45
bangmal
dalekdukesboy
g200 fail? lol...I think that's only applicable at it's original 500+ price tag, but once the price cuts came into effect I can hardly say that the g200 isn't a good gpu...considering it takes 2 of ati's best cpus on one card (4870x2) to beat it...
takes 2 gpu? not really...
http://www.anandtech.com/video/showdoc.aspx?i=3415&p=7

and oh yes, G200 are utter failures, for the pathetic performance with transistors and die size like those
Posted on Reply
#46
TheGuruStud
bangmal
takes 2 gpu? not really...
http://www.anandtech.com/video/showdoc.aspx?i=3415&p=7

and oh yes, G200 are utter failures, for the pathetic performance with transistors and die size like those
I can't believe you just linked to anand. :shadedshu
That's sad.

And BTW, if that's real, apparently you missed the other 5,000 games that the geforces excel at.

And at that res (that no one will play at or use AA), it just makes it even more irrelevant.
Posted on Reply
#47
Wile E
Power User
eidairaman1
and there are naysayers of deneb requiring 1.6 is too much, pfffttt, this just proves those people wrong, especially when intel specified anything higher than 1.6 for their CPUs is too much and will kill them quick.
1.6V will kill a Deneb too, mark my words. ;) And people were saying Deneb needing 1.6V to reach 4Ghz is too much because of the Deneb FX 4GHz at stock rumor, meaning it's too much voltage for them to be able to release a 4GHz chip from the factory.

Besides, this Nehalem is running at 4.2GHz, it's altogether possible it only takes 1.6V to hit 4GHz. It takes huge voltage increases to get a few MHz when you are reaching the limits of a cpu. I'm willing to bet a 4Ghz Nehalem will easily outpace a 4Ghz Deneb in just about everything.

But at any rate, like I said in another thread, if this is indicative of what we can expect from the retail chips, it looks like I'll be sticking with my QX9650. I'll just pick up 2 new DDR3 boards(one Xfire and one SLI board), a good controller card, some DDR3, and ride it out until 1366 cpus clock better. Under DICE cooling, the QX9650 should be able to clock far enough to overtake the Nehalems.
Posted on Reply
#48
eidairaman1
The Exiled Airman
Well as its said, Nehalem is more of a Server System upgrade than an Enthusiast or average user upgrade.
Wile E
1.6V will kill a Deneb too, mark my words. ;) And people were saying Deneb needing 1.6V to reach 4Ghz is too much because of the Deneb FX 4GHz at stock rumor, meaning it's too much voltage for them to be able to release a 4GHz chip from the factory.

Besides, this Nehalem is running at 4.2GHz, it's altogether possible it only takes 1.6V to hit 4GHz. It takes huge voltage increases to get a few MHz when you are reaching the limits of a cpu. I'm willing to bet a 4Ghz Nehalem will easily outpace a 4Ghz Deneb in just about everything.

But at any rate, like I said in another thread, if this is indicative of what we can expect from the retail chips, it looks like I'll be sticking with my QX9650. I'll just pick up 2 new DDR3 boards(one Xfire and one SLI board), a good controller card, some DDR3, and ride it out until 1366 cpus clock better. Under DICE cooling, the QX9650 should be able to clock far enough to overtake the Nehalems.
Posted on Reply
#49
largon
1.7v?
*cough*

Obviously, CPU-Z has no clue about the actual vCore...
*pokes people with a pointy stick*
Posted on Reply
Add your own comment