Early Nehalem (Core i7) Benchmarks Show Off Performance Gains

[840quadra]

Last I spoke with a friend that works at a local apple store he was telling me to wait till January for an updated macbook with the new i7. He probably doesn't have any better idea than anyone else so I'm not waiting. Plus I could wait for that and then 6 month later something better will come along, sometimes you just have to jump in

You are correct. Even with an Apple Employee NDA (yes the store employees sign this too), they still don't know anything before we do (most cases). Apple is aware those employees are most likely to leak information to the general public.

I am waiting for a new Nehalem system too, however I am going to happily use my new dual graphics MBP until those models are released.

 

[tenks]

These two are the only two I wondered about. I was under the impression(from Fudzilla) that Nvidia has QPI license, which would mean they can make chipsets for Nehalem. I also was under the impression that FB-DIMM was only for the MP servers not the DP systems. Registered DDR 3 was going to be used in Gainestown.

I edited my original post to be a bit more specific about the Nvidia license. Nvidia said they had a "QPI" license because of 2 reasons. QPI spelled the death of 3rd party chipset manufarctures for Intel parts: ATI, Nvidia, VIA, SIS etc. The memory controller has been integrated on die and the FSB is no longer there. Why would you need a 3rd party chipset when there is hardly anything on the chip that could change or improve from Inte's own? There were also a TON of rumors of Nvidia exiting the chipset market (which is partly true, their chipset for core2duo will be the last DESKTOP intel chipset from them for awhile) altogether. By making this announcement its lets the press and its shareholders know it still can make profit off the Nehalem train w/o actually making a phsyical product, ie chipset. Their QPI license is so they can run SLI natively on the Nehalem platform, which will be a cash cow for them. Well at least for now, thats all they have access too. Intel and Nvidia have a strange relationship. Who knows...

Gainestown might have a DDR3 variant and an FB-DIMM variant. Gainestown and Beckton are socket compatible meaing they are essentially the same chip. Ones just meant for dual set ups and one is used for 2+. So with that being said, they both can USE FB-DIMMs, at least theoretically...


Also for those who have some time are REALLY interested in why Nehalem is such a big deal, please read Anandtech's review: http://www.anandtech.com/cpuchipsets/intel/showdoc.aspx?i=3448

Ive been reading his stuff for over 10 years and to this day cant get a more complete unbiased review then his. GREAT READ

 

[Eidorian]

Using desktop chips? Didn't think they had the pins required to communicate across the sockets. It would be pretty slick though. The interesting part is the desktop chip (4 core, 8 lp's) being able to keep up with the 8 core server (DP) variant. May be Apple could come out with a MP Lite using that chip till the DP chips are available, then bust out with the 8 core 16 lp model. They could even force Registered DDR3 for the desktop version so the RAM can be moved across (FBDIMM is only for the MP systems).

Why use special pin-outs and SMP features only found on Xeon processors when you can use QPI? Nehalem multi-socket isn't the same game like it was in the past with the front side bus. Gainestown and Bloomfield are using the same LGA1366 socket as well.

 

[Umbongo]

I edited my original post to be a bit more specific about the Nvidia license. Nvidia said they had a "QPI" license because of 2 reasons. QPI spelled the death of 3rd party chipset manufarctures for Intel parts: ATI, Nvidia, VIA, SIS etc. The memory controller has been integrated on die and the FSB is no longer there. Why would you need a 3rd party chipset when there is hardly anything on the chip that could change or improve from Inte's own? There were also a TON of rumors of Nvidia exiting the chipset market (which is partly true, their chipset for core2duo will be the last DESKTOP intel chipset from them for awhile) altogether. By making this announcement its lets the press and its shareholders know it still can make profit off the Nehalem train w/o actually making a phsyical product, ie chipset. Their QPI license is so they can run SLI natively on the Nehalem platform, which will be a cash cow for them. Well at least for now, thats all they have access too. Intel and Nvidia have a strange relationship. Who knows...

Gainestown might have a DDR3 variant and an FB-DIMM variant. Gainestown and Beckton are socket compatible meaing they are essentially the same chip. Ones just meant for dual set ups and one is used for 2+. So with that being said, they both can USE FB-DIMMs, at least theoretically...

You've got some of your information wrong. Gainestown and Bloomfield are the same socket and both use DDR3. Though I doubt you will be able to use either in boards intended for the other. Beckton uses a different socket and will require FB-DIMMs.

 

[tenks]

You've got some of your information wrong. Gainestown and Bloomfield are the same socket and both use DDR3. Though I doubt you will be able to use either in boards intended for the other. Beckton uses a different socket and will require FB-DIMMs.

Oops sorry, that I did. My humble apologies. I had the right idea, just the wrong codenames. I thought the Bloomies were going to be the only 1366s and there was going to be 2 versions of whatever socket Beckton uses. Thanks for straightening that up.

 

[diamond.g]

Oops sorry, that I did. My humble apologies. I had the right idea, just the wrong codenames. I thought the Bloomies were going to be the only 1366s and there was going to be 2 versions of whatever socket Beckton uses. Thanks for straightening that up.

Sadly, I didn't even notice the mixup.

Why use special pin-outs and SMP features only found on Xeon processors when you can use QPI? Nehalem multi-socket isn't the same game like it was in the past with the front side bus. Gainestown and Bloomfield are using the same LGA1366 socket as well.

Hmm, according to Wiki the Bloomfield only has 1 QPI whereas Gainestown has 2. So I wonder if the Bloomsfield chip will work in a DP board even though it only has 1 QPI.

 

[Eidorian]

Hmm, according to Wiki the Bloomfield only has 1 QPI whereas Gainestown has 2. So I wonder if the Bloomsfield chip will work in a DP board even though it only has 1 QPI.

I'd wait for final Xeon boards and Nehalem Skulltrail to come out. It's still a bit murky right now without final product.

You are correct that Gainestown does have 2 QPI. It's still the same socket and RAM support as Bloomfield though.

 

[diamond.g]

I'd wait for final Xeon boards and Nehalem Skulltrail to come out. It's still a bit murky right now without final product.

You are correct that Gainestown does have 2 QPI. It's still the same socket and RAM support as Bloomfield though.

Yeah, really it is more trying to figure out if Bloomfield could be used in the MacPro till Gainestown was released.

 

[tenks]

Why use special pin-outs and SMP features only found on Xeon processors when you can use QPI? Nehalem multi-socket isn't the same game like it was in the past with the front side bus. Gainestown and Bloomfield are using the same LGA1366 socket as well.

I could be wrong, but what I think he was trying to say is, because there is only one active QPI link in Bloomies, the pins for the 2nd QPI link are disabled. Just because there is QPI doesnt mean you can go dual. You need 2 QPI's to have a correct DP setup with Nehalem, which means Bloomies probably cant be plugged into a dual socket board. Which answers that question.. This is why gainestown exists....

Also reading through Anand's review, he again reaffirms the Nvidia "QPI" license
[
"The biggest feature of X58 is that with proper "certification" by NVIDIA, motherboard makers can include support for the right BIOS flags to allow NVIDIA's drivers to enable SLI on the platform. Meaning the X58 will be the first Intel chipset to support both CrossFire and SLI multi-GPU solutions without the use of any NVIDIA silicon. There's a per-motherboard fee from NVIDIA for each certified X58 board sold and thus not all boards will be certified, the most prominent of which is Intel's own X58 board. "

Cash cow for nvidia like I said...So all the rumors are actually true I guess...They arent making any chipsets (yet) for i7 but they are still making money off of it. Sneaky...

 

[Eidorian]

I could be wrong, but what I think he was trying to say is, because there is only one active QPI link in Bloomies, the pins for the 2nd QPI link are disabled. Just because there is QPI doesnt mean you can go dual. You need 2 QPI's to have a correct DP setup with Nehalem, which means Bloomies probably cant be plugged into a dual socket board. Which answers that question.. This is why gainestown exists....

Also reading through Anand's review, he again reaffirms the Nvidia "QPI" license

"The biggest feature of X58 is that with proper "certification" by NVIDIA, motherboard makers can include support for the right BIOS flags to allow NVIDIA's drivers to enable SLI on the platform. Meaning the X58 will be the first Intel chipset to support both CrossFire and SLI multi-GPU solutions without the use of any NVIDIA silicon. There's a per-motherboard fee from NVIDIA for each certified X58 board sold and thus not all boards will be certified, the most prominent of which is Intel's own X58 board. "
"

Nehalem Skulltrail is going to still use X58 though. You also forgot something else...

The biggest feature of X58 is that with proper "certification" by NVIDIA, motherboard makers can include support for the right BIOS flags to allow NVIDIA's drivers to enable SLI on the platform. Meaning the X58 will be the first Intel chipset to support both CrossFire and SLI multi-GPU solutions without the use of any NVIDIA silicon. There's a per-motherboard fee from NVIDIA for each certified X58 board sold and thus not all boards will be certified, the most prominent of which is Intel's own X58 board. Luckily we also had access to ASUS' P6T Deluxe which is certified, giving us the ability to look at CrossFire and SLI scaling on X58 vs. other platforms.

 

[tenks]

Uh ok? Im kinda lost at what you're trying to say?

I never said Nehalem wasnt going to use x58 for skulltrail or anything related?

And as far as Intel's own boards not being SLI licensed..Thats not that big of a deal. The people who want SLI usually go for the ASUS, Gigabyte, DFI etc boards. Not intel's.

 

[kntgsp]

Jesus christ, welcome to 3 months ago.

The benchmarks for these has been around forever.

 

[mdriftmeyer]

confirms what i thought. nice speed increase but not revolutionary. in day to day computing 10%-20% increase is barely noticable.

The upgrade will only be useful for Mac pro owners with professional projects where time is money.

When Nehalem is out for notebooks the speed increase for most users is negligible. The jump form dual to quadcore will be much more important.

It's always the same: the hype from core duo to core2duo, from core2duo to penryn, from penryn to nehalem is always huge, the real world benefit for office work, fun and similar uses is rather limited.

Other factors like battery life, GPU, HD speed should be more important for the average user. No need to delay a purchase.

The increase won't be truly noticed until Snow Leopard 10.6 and true Multi-core leverage throughout the OS; and this change will effect all current multi-core CPU systems.

 

[iSamurai]

well this Core i7 thing is new to me, and so I guess it won't be making into the laptops anytime soon. Wikipedia didn't explain it nicely

Anyone know what Apple would use this chip in the future for? Mac Pro? Dunno about the higher-end iMacs because aren't they really laptop-like (internals)?

 

[Eidorian]

The increase won't be truly noticed until Snow Leopard 10.6 and true Multi-core leverage throughout the OS; and this change will effect all current multi-core CPU systems.

I noticed an improvement in Windows XP moving from my old E2160 @ 2.4 GHz to the Q6600...

Snow Leopard this and that is getting rather old. People make it seem like multiple processors and cores aren't of any benefit at all right now.

 

[Umbongo]

well this Core i7 thing is new to me, and so I guess it won't be making into the laptops anytime soon. Wikipedia didn't explain it nicely

Anyone know what Apple would use this chip in the future for? Mac Pro? Dunno about the higher-end iMacs because aren't they really laptop-like (internals)?

The Mac Pro is likely to be the first Nehalem Mac. In the future Nehalem will replace all the current processors so it will be in everything.

 

[powers74]

i7

Oh.......

I thought these things were supposed to be insanely faster. Guess it doesn't really matter when I get a Pro then.

 

[albusseverus]

Anyone remember the 'old' days when the rule of thumb was that you needed 400% speed improvement for it to be noticeable in real world usage?

I understand that upgrades come faster now, but 20%, 40%, even 100% doesn't make much difference.

Maybe in the Intel world, it's important to prove that speed hasn't decreased with every incremental change !

 

[iMacmatician]

well this Core i7 thing is new to me, and so I guess it won't be making into the laptops anytime soon. Wikipedia didn't explain it nicely

Anyone know what Apple would use this chip in the future for? Mac Pro? Dunno about the higher-end iMacs because aren't they really laptop-like (internals)?

Given the current line up, these are the possible CPUs for the Mac models:

Beckton - None
Gainestown - Mac Pro, Xserve (H1 2009)
Bloomfield - None
Lynnfield - None
Clarksfield - iMac, MacBook Pro (Q3/Q4 2009)
Havendale - None
Auburndale - iMac, MacBook Pro, MacBook (Q1 2010)

Note that Apple might restrict certain Mac lines to certain CPUs.

 

[arobasefr]

Nehalem in the "MAC"

)
Bloomfield - None
.

I guess the first benchmarks of nehalem on a hackintosh will convince a lot of geeks not to wait for Q3 2009.!

And I a don't even speak about Snow leopard on it !

 

[diamond.g]

Oh.......

I thought these things were supposed to be insanely faster. Guess it doesn't really matter when I get a Pro then.

Supposedly in encoding video the 4 core rig kept up with the 8 core rig. If that isn't insane for you I dunno what else is.

Given the current line up, these are the possible CPUs for the Mac models:

Beckton - None
Gainestown - Mac Pro, Xserve (H1 2009)
Bloomfield - None
Lynnfield - None
Clarksfield - iMac, MacBook Pro (Q3/Q4 2009)
Havendale - None
Auburndale - iMac, MacBook Pro, MacBook (Q1 2010)

Note that Apple might restrict certain Mac lines to certain CPUs.

It would be sweet if Apple used Bloomfield in the interim for the Mac Pro till Gainestown is ready.

 

[NewSc2]

*yawn* going from 4-core to 8-core was an improvement enough. 20% is nice, but that should be expected. Not many people truly care about a ~10% speed boost.

 

[diamond.g]

*yawn* going from 4-core to 8-core was an improvement enough. 20% is nice, but that should be expected. Not many people truly care about a ~10% speed boost.

We're not quite ready to release a full set of test results, but trust me when I say this thing is fast. How fast? Fast enough that a one-socket Nehalem system armed with a top-end, $999 Core i7-965 could be an extremely cost-effective solution for anyone in the market for a dual-socket Xeon system outfitted with two quad-core processors. I've a feeling that the numerous processor enhancements Intel built into Nehalem would make the chip potent in its own right, but combined with the platform upgrades, it's the rarest of rare gems—a $1,000 CPU that might be worth buying.

Nehalem Article

From the looks of it, high end Core i7 should be approaching par with the DP Xeons. Which to me is nuts.

 

[3D-Troll]

Oh and is there a Dual-Link-DVI to 2 x DVI-I adapter out there? I wanna hook up two 20" 1680x1050 displays to a unibody Macbook.

Try the Matrox DualHead2Go, this should do the trick.

Steffen

 

[Firefly2002]

Meh. I had thought improvements were generally 40-50%. 20-25ish leaves me pretty underwhelmed.