Intel Releases Core i7 ''Sandy Bridge-E'' Processors

[btarunr]

Intel today released its Core i7-3000 series processor family, codenamed "Sandy Bridge-E". These new processors, along with the new Intel X79 chipset, make up for an entirely new platform. The processors are an upscale of the Sandy Bridge architecture found on chips in the LGA1155 package. The Sandy Bridge-E silicon measures 20.8 x 20.9 mm, with a humungous transistor count of 2.27 billion. In its Core i7-3000 configuration, the silicon has up to 6 cores, up to 15 MB of L3 cache, four DDR3 memory channels, and 40 PCI-Express 3.0 lanes ("some" devices "may" support Gen 3.0, Intel's words).

Sandy Bridge-E has the same instruction set as Sandy Bridge, which includes SSE up to version 4.2, AVX, AES, and features Turbo Boost 2.0, HyperThreading. It's the memory controller that's complete upscale. It features four independent 64-bit paths to DDR3 DIMMs, making it a quad-channel DDR3 IMC. DDR3-1600 MHz is natively supported. There are three models, the Core i7-3960X Extreme Edition leads the pack with a clock speed of 3.30 GHz, 3.90 GHz top Turbo Boost speed, and 15 MB of L3 cache. It has 6 cores and 12 threads with HTT enabled. This chip has all its multipliers unlocked and is geared for overclocking. It is priced at US $990 in 1000 unit tray quantities, though retailers might draw a decent margin for the boxed parts.



The next best chip in the series is Core i7-3930K. With clock speeds of 3.20 GHz and 3.80 GHz (Turbo), this chip has a slightly smaller L3 cache size of 12 MB, though it is still unlocked and geared for overclocking. Like the i7-3960X, this is a 6 core / 12 thread chip. This chip commands a price of $555. Touted to be the most affordable model, the Core i7-3820 is a quad-core part drawn out of disabling two cores (there's no evidence so far that they can be unlocked). With HTT enabled, this chip offers 8 threads. Its L3 cache is further reduced, to 10 MB (still higher than any preceding Core i7 quad-core model). Unfortunately, this chip is "partially unlocked", meaning that its base clock multiplier is locked, though you can still effectively overclock it by tinkering with the base clock. What's even more depressing is that this chip won't be available until Q1 2012. It is supposed to be priced in the $299~$399 range. This means that the only people building Sandy Bridge-E desktops this Christmas will be the ones with at least $600 to spare for a processor.

Moving on to the platform itself, the processor is built on the new LGA2011 package, it's the largest CPU package by dimensions, in recent times. Over its 2011 pins, the processor gives out four DDR3 memory channels and 40 PCI-Express 3.0 lanes, a DMI 4 GB/s connection to the X79 chipset, and a large number of pins handling power. The X79 chipset itself doesn't differ much from the P67 chipset in terms of the kind of connectivity it offers, except support for Intel Smart Response SSD-caching technology.

[mcloughj]

Expensive. No surprise there but I think a 2600K (or 2500) is still the way to go for most gamers.

[btarunr]

Originally Posted by mcloughj

Expensive. No surprise there but I think a 2600K (or 2500) is still the way to go for most gamers.

Yup. Not just gamers, but pretty much anyone who doesn't encode videos for a living.

[n-ster]

NOOOOOOOOOOOO MY QUADDDDDDDDDD

Damn I was looking forward to X79... Guess I'll have to wait? Damn this sucks ass

But prices are pretty good so far

[buggalugs]

Yep, major disappointment.

Anand has a review up

http://www.anandtech.com/show/5091/i...high-end-alive

Fuck you Intel

[qubit]

I'm mighty disappointed with this product - same gaming performance as SB. This is what happens when Intel don't have any head to head competition.

I made more noises about it here.

[btarunr]

So W1zzard was bang on about the front-page poll options, after all.

[n-ster]

I hope MC offers that i7 3930K for cheap by the time New Year's has past... So far it is priced at first day launch prices lol

http://www.microcenter.com/single_pr...uct_id=0376493

[RejZoR]

Only good thing about it is AVX instruction. Which is also avaiable on 2600K as far as i know. So who cares really.

[qubit]

Originally Posted by btarunr

So W1zzard was bang on about the front-page poll options, after all.

W1zz is always right. What are you trying to say?! qubit dons tinfoil hat

[xenocide]

Performs top of the line in every benchmark--major disappointment.

[n-ster]

http://www.ncix.com/products/?sku=65...SI%2FMicroStar

http://www.ncix.com/search/?categoryid=0&q=x79

259.99$, 279.99$, 299.99$... This is in CANADA where prices are usually a rip-off... and on LAUNCH DATE. Hope to see 250$ or less in the future

[xenocide]

Originally Posted by qubit

W1zz is always right. What are you trying to say?! qubit dons tinfoil hat

I especially like that he knew the FX-8150 wouldn't be able to compete head to head with the i7-2600k in the Spring when everyone was convinced it would, and in most cases even best it.

[Rowsol]

Originally Posted by qubit

I'm mighty disappointed with this product - same gaming performance as SB.

There wasn't any reason to think it would be any better, unless you know of a game that uses 12 threads.

This is intel's bulldozer.

[qubit]

Originally Posted by Rowsol

There wasn't any reason to think it would be any better, unless you know of a game that uses 12 threads.

This is intel's bulldozer.

No, it isn't Bulldozer, lol, but it's what happens when you have Bulldozer 'compete' against Intel's current lineup. Check out the other post I linked to, which explains my point in more detail.

[jimmyxxx]

40 PCI-Express 3.0 lanes

Typo there, or am i wrong?

[LAN_deRf_HA]

Why is everyone being such a dumbass about this. 1366/1156 had the same gaming performance too. There was never any illusions otherwise. If you're disappointed it's your own fault for having shit for memory.

[n-ster]

Originally Posted by jimmyxxx

Typo there, or am i wrong?

No, this is exactly it and known for a while... Like one 16x PCI-E 3.0 and three 8x PCI-E 3.0 or two 16 one 8 etc etc

[Live OR Die]

Originally Posted by mcloughj

Expensive. No surprise there but I think a 2600K (or 2500) is still the way to go for most gamers.

Hell no im going to go all out get me one of these i7-3960X Windows should load support fast , Did any one think intel wanted this chip out this year because its the year 2011 :P

[Benetanegia]

I dn't understand the dissapoinment honestly. It's been well known that this was only going to be SB with 6 cores, so single core and light threaded performance was going to be the same. This includes gaming, since no game trully uses much more than 4 cores. Those which do use more threads (few and far between) are more or less taken care of by hyperthreading. Intel can't just create more gaming performance out of nowhere when games don't use more resources.

Heavily multi-threaded apps' performances are were they should be, at a 40%-50% increase over 2600k.

Intel did make a mistake if we look at it from a marketing pov, and they made it several months ago by releasing the midrange part first. Mid-range, in this consolized gaming market and lazy consumer app programers will always take the cake, specially when the difference is the number of cores. That's why everyone always try to release high-end first, even on GPUs which are always used to almost 100% of their potential.

[n-ster]

Originally Posted by Live OR Die

Hell no im going to go all out get me one of these i7-3960X Windows should load support fast , Did any one think intel wanted this chip out this year because its the year 2011 :P

I think that was the whole point when they announced it lol... If it were for 2012 they would have introduced an extra pin somehow lol

[CDdude55]

HardOCP's review is up: http://hardocp.com/article/2011/11/1...essor_review/1

Nothing too surprising, in multithreaded applications it takes off, in single-threaded application it's pretty terrible and the older Sandy Bridge come out ahead in a few instances.

When it comes to gaming it looks to be another Bulldozer, it's pretty underwhelming. It performs the same or a little worse in pretty much all titles against older chips.

The only time this chip is a good buy is if you have a lot of money to spend and you do a lot of heavy multithreaded work, that's where it seems to shine.(not surprisingly)

[claylomax]

Originally Posted by qubit

W1zz is always right. What are you trying to say?! qubit dons tinfoil hat

Still he'll change his test bench: http://www.techpowerup.com/forums/sh...11&postcount=7

[n-ster]

It was never supposed to see an improvement in games though.... Maybe in multi-GPU though

[brandonwh64]

I think this is more for server/heavy encoder systems