Friday, July 22nd 2011

Core i7-3960X About 47% Faster On Average Than Core i7-990X: Intel

Slides of a key presentation to Intel's partners was leaked to sections of the media, which reveal Intel's own performance testing of the Core i7-3960X Extreme Edition, the top-model of the socket LGA2011 "Sandy Bridge-E" processor series. Meet the family here. In its comparison, Intel maintained the Core i7-990X Extreme Edition socket LGA1366 processor as this generation's top offering. It was pitted against the Core i7-3960X in a battery of tests that included some enthusiast favourites such as Cinebench 11.5, POV-Ray 3.7, 3DMark 11 physics, Pro-Show Gold 4.5, and some OEM favourites such as SPECint_rate base2006, SPECfp_rate base2006, and SiSoft SANDRA 2011B multimedia and memory bandwidth.

From these test results, the Core i7-3960X Extreme Edition is pitched to be about 47.25% faster on average, compared to Core i7-990X Extreme Edition. Intel is attributing the performance boost, apart from the normal IPC increase, to the 33% higher bandwidth thanks to the quad-channel DDR3 IMC, and the new AVX instruction set that accelerates math-heavy tasks such as encoding. The Core i7-3960X Extreme Edition is an upcoming socket LGA2011 six-core processor that is clocked at 3.30 GHz, with Turbo Boost speed of up to 3.90 GHz, with 12 threads enabled by HyperThreading technology, and 15 MB L3 cache. It will release by either late 2011 or early 2012.
Source: Donanim Haber
Add your own comment

116 Comments on Core i7-3960X About 47% Faster On Average Than Core i7-990X: Intel

#26
mlee49
btarunrIt "is up to" 111% faster. My statement included the term "on average", given the provided test results.
Still proves the point that the marketing percentages are wack.
Posted on Reply
#27
[H]@RD5TUFF
mlee49Still proves the point that the marketing percentages are wack.
I don't think anyone would question that, but both companies (AMD and Intel) use questionable math to inflate numbers to make X look like it's 50+% faster than it's competitors y chip, this is nothing new.
Posted on Reply
#28
Pestilence
Meh.... 6 cores is so 2008. Intel should have been ballsy and went with the 8 core

Im staying 1155 and just going Ivy Bridge
Posted on Reply
#29
[H]@RD5TUFF
PestilenceMeh.... 6 cores is so 2008. Intel should have been ballsy and went with the 8 core
While I agree, I will reserve judgement until I see hard benchmark numbers.
Posted on Reply
#30
FordGT90Concept
"I go fast!1!11!1!"
mlee49Realistic interpretation:

A cherry picked 3960X LGA2011 is up to 47% faster than a stock 990x LGA1366.
I doubt it was "cherry picked" (and not that it matters, they didn't overclock it). They probably just grabbed a hexa-core CPU off the line to compare to the previous generation of hexa-cores. Remember, an octo-core w/ hyperthreading is being developed too. Intel is confident they have the performance crown so there is no reason for them to work the numbers.
Posted on Reply
#31
Velvet Wafer
happitaThat's true, but done right, cache can also increase performance as well. Leakage is the main issue when increasing the cache.
of course it will increase performance, without that it would be pointless at all!:D
but the amount of problems, more cache generates, exceeds the benefits by a lot;)
Posted on Reply
#32
mlee49
FordGT90ConceptI doubt it was "cherry picked" (and not that it matters, they didn't overclock it). They probably just grabbed a hexa-core CPU off the line to compare to the previous generation of hexa-cores. Remember, an octo-core w/ hyperthreading is being developed too. Intel is confident they have the performance crown so there is no reason for them to work the numbers.
I'm standing firm on it being a cherry picked chip, it cant possibly be retail and only an ES leaked will be dispute enough for me. [Chiphell better rebuttle]
Posted on Reply
#33
kid41212003
No cherrypick shit! You don't need a cherrypicked chip to run default speed!

Turbo mode:

i7 990X = 3.73GHz
i7 3960X = 3.9GHz

It's quite obvious.
Posted on Reply
#34
dumo
Cherry picked 3960X should be able to run 5900Mhz+...means bootable @ 59X multi
Posted on Reply
#35
FordGT90Concept
"I go fast!1!11!1!"
mlee49I'm standing firm on it being a cherry picked chip, it cant possibly be retail and only an ES leaked will be dispute enough for me. [Chiphell better rebuttle]
Intel wouldn't publish numbers on a named chip if it were an engineering sample. I'd bet they got the Core i7-3960X and other chips ready to ship out the door but they're waiting for motherboard manufactuers to get their ducks in a row before they launch.
Posted on Reply
#36
[H]@RD5TUFF
FordGT90ConceptIntel wouldn't publish numbers on a named chip if it were an engineering sample. I'd bet they got the Core i7-3960X and other chips ready to ship out the door but they're waiting for motherboard manufactuers to get their ducks in a row before they launch.
I hope this is true, I know intel is on a spending spree buying up all kinds of test handlers, at work we just sold them 35 units, and they are asking for more.
Posted on Reply
#37
Assimilator
It's as if AMD's Bulldozer architecture cried out in terror, and was suddenly silenced.

By a curbstomp.

From a mech.
Posted on Reply
#38
Aevum
FFS... Was the new socket an absolute necesity.

Im getting tierd with this. AMD has 2 active sockets for the entire range. FM1 and AM3+.

Intel currently has 3 active sockets and a 4th on the way. 1366 1155 1156 and now 2011.

And thats without mentioning the chipset dance. even if the socket is the same the chipset has to be the correct one...

Please intel. pick 1 socket and make the damn chipsets compatible.
Posted on Reply
#39
[H]@RD5TUFF
AevumFFS... Was the new socket an absolute necesity.

Im getting tierd with this. AMD has 2 active sockets for the entire range. FM1 and AM3+.

Intel currently has 3 active sockets and a 4th on the way. 1366 1155 1156 and now 2011.

And thats without mentioning the chipset dance. even if the socket is the same the chipset has to be the correct one...

Please intel. pick 1 socket and make the damn chipsets compatible.
Well then your going to be mad at AMD as they are phasing out AM3+ come the end of 2012 for another one.
Posted on Reply
#40
FordGT90Concept
"I go fast!1!11!1!"
Recycling old sockets holds them back. Just look at AM2 to AM3+. What changed? DDR3, and that's about it. Look at Intel on the other hand and every new socket brought major changes. LGA1366 = tri-channel memory, LGA2011 = quad-channel memory, LGA1156 = DDR3 RAM, LGA1155 = GPU in chip, etc.

AMD fell into an advancement hole on AM2 so they had no real reason to push out different sockets like Intel did.
Posted on Reply
#41
DrunkenMafia
Wow, Quad channel DDR3!!

This would have to over $2k when it comes out, surely.
Posted on Reply
#42
ensabrenoir
AssimilatorIt's as if AMD's Bulldozer architecture cried out in terror, and was suddenly silenced.

By a curbstomp.

From a mech.
Thou art an poet an poet amongst druken sailors:roll:
Posted on Reply
#43
Steven B
this site(not TPU) has been wrong before, but i do hope they are right.
Posted on Reply
#44
WhiteLotus
I am so fucking confused over what is what.
i7, seemingly arbitrary numbers, and then a letter.

Is that a quad, a duo, a what? Is it good, is it better, does it make toast?!? And AMD are guilty too. Can I please just get a decent scale over what the fuck is what.

Oh and then there's the whole some iX use this socket, some use that. Can they just use ONE socket for ONE "species" of chip.
Posted on Reply
#45
Crap Daddy
In a way it's sad that Intel is in a category of its own. At this level there is no competition so they will sit down and think how much to charge for this, based on whatever numbers that Italian name dude who's running the shop has in mind.
Posted on Reply
#46
sno.lcn
AevumFFS... Was the new socket an absolute necesity.

Im getting tierd with this. AMD has 2 active sockets for the entire range. FM1 and AM3+.

Intel currently has 3 active sockets and a 4th on the way. 1366 1155 1156 and now 2011.

And thats without mentioning the chipset dance. even if the socket is the same the chipset has to be the correct one...

Please intel. pick 1 socket and make the damn chipsets compatible.
Are new CPU models still coming out for s1156? Will any new ones be coming out for s1366 when s2011 is released? :confused:

Anyway, looking at SB and SB-E side by side, does putting them both in the same package really sound reasonable? No way :p
DrunkenMafiaWow, Quad channel DDR3!!

This would have to over $2k when it comes out, surely.
Extreme series CPUs from Intel have always been in a certain price range, don't expect this to deviate too drastically from that.
dumoCherry picked 3960X should be able to run 5900Mhz+...means bootable @ 59X multi
Are we already assuming these clock just like sandy bitch chips? :D
Posted on Reply
#47
Pestilence
sno.lcnAre new CPU models still coming out for s1156? Will any new ones be coming out for s1366 when s2011 is released? :confused:
Sandy Bridge uses Socket 1155... 1156 was Lynnfield.

There's been a rumor floating around XS that was first suggested over at [DH] that intel will drop prices on the 2500K/2600K from 219.99 to 199.99 and 314.99 to 289.99 to make way for a 2800K that will be 6 cores with no HT for 349.99 but this is PURELY speculation and i haven't seen any ES of it.
mlee49A cherry picked 3960X LGA2011 is up to 47% faster than a stock 990x LGA1366.
Why would it be cherry picked? SB-E is already on B2 steppings. Hell C0 will probably be retail.
Posted on Reply
#48
Steven B
honestly the change is socket is probably why Intel has better performance, they aren't limited by the socket/chipset.
Posted on Reply
#49
Pestilence
Steven Bhonestly the change is socket is probably why Intel has better performance, they aren't limited by the socket/chipset.
Amd wasn't limited by the sockets. They just decided to stay with AM2 to appease the masses and look where it has gotten them.
Posted on Reply
#50
Steven B
yea i am sure your right lol. Its either that, or its becuase their revenue is so much less than Intel that they don't bother, or can't.

And yes you can be limited by a socket, by its # of pins just for vcc and ground for a CPU with so much more TDP is a big deal, add a few more cores, some PCI-E controller, IMC to quadchannel, and so on.

if you notice with LGA1366 intel has a IOH(northbridge) and ICH(southbridge), with LGA2011 and LGA1155 there is only a PCH, like the A75 chipset(called FCH) if you are familiar with AMD.

So AMD is still using a two part chipset, while Intel has moved to all single chipset. These moves, such as the testing of this with LGA1156, and then sticking with two for LGA1366, shows Intel has the ability to heavily modify the substructure other than the cores, while with lets say Phenom 2 they couldn't add an onboard PCI-E controller, they are limited to the NB, same with BD, they have to use that NB, they are also limited to power package, and thus they came out with true AM3+ boards, AM3 board will be limited and so on.

Its a lot of confusion, Intel doesn't bring out new sockets every year.
Posted on Reply
Add your own comment
Apr 26th, 2024 09:47 EDT change timezone

New Forum Posts

Popular Reviews

Controversial News Posts