Curse

Evil Genius · Nov 03, 2008, 06:25 AM // 06:25

What is Core i7?
Core i7 is Intel's latest family of desktop processors: the sucessor to the Core 2 family and built on the new Nehalem microarchitecture.

Will it work in my current system?
Core i7 requires new mothboards based on the X58 chipset and DDR3 RAM.

Why is Core i7 important?
Core i7 incorporates the most signifanct architecture changes in Intel's CPUs since the days of Pentium Pro in 1995. There will be a new socket, new CPUs, new chipset, new memory architecture and new overclocking methods.

The chips

Note the prices are in USD$ per 1000 units. Prices in stores will be greater.

Intel Core i7 940, 920 and 965 Extreme Edition Reviews
AnandTech
PC Perspective
Hot Hardware
Legion Hardware
TweakTown
t-break
Driver Heaven
Trusted Reviews
Benchmark Reviews
Techgage
TechSpot
Guru3D [1] [2]

I encourage you to read a review if interested, but for those who want a quick guage of performance, this is the only summary image I could find. It is from the OCAU review:

Motherboards
ASUS P6T Deluxe OC Palm Edition
EVGA X58 SLI
Gigabyte GA-EX58-UD5P X58
ASUS Rampage II Extreme X58

ECS X58B-A
ASUS Rampage II Extreme X58

DDR3 Memory
Kingston HyperX DDR3 3GB 2GHz Triple-Channel
G.Skill DDR3 1600 Triple Channel
Kingston HyperX Triple Channel DDR3 2000

Brianna · Nov 03, 2008, 05:16 PM // 17:16

Nice, thanks for the post - I didn't even know there was info about this until now.

Also don't forget this one: http://techreport.com/articles.x/15818

And this too maybe: http://techreport.com/articles.x/15816

I'm reading them, very long but if you aren't doing anything it's manageable.

Quaker · Nov 03, 2008, 05:44 PM // 17:44

Is Guru getting any ad revenue from this? They should.

Anyway, it just confirms what we all suspected - bleeding edge performance at a bleeding edge price, with little effect on gaming performance.
It will be interested to see how the low end one compares to other similarly priced choices.

Cyb3r · Nov 04, 2008, 07:41 PM // 19:41

Little effect on gaming quaker ya seen the Scores in the guru3d article with Sli compared to the old ones?

It's a 50% increase in performance on the set they used....

Snograt · Nov 04, 2008, 08:16 PM // 20:16

Nah, won't bother - it's slower in Quake 4

Admael · Nov 04, 2008, 09:02 PM // 21:02

I assume the "new overclocking methods" is referring to this?

http://www.theinquirer.net/gb/inquir...damage-nehalem

deluxe · Nov 04, 2008, 09:26 PM // 21:26

I'm not impressed with performance in current days games. My Q6600 @ 3.8 still outperforms these new cpu's.
Not worth the money yet to start upgrading now. DDR3 currently also isnt much better than the latest low latency DDR2.

gaming performance

Cyb3r · Nov 04, 2008, 09:29 PM // 21:29

Haha good one snograt

Evil Genius · Nov 04, 2008, 10:21 PM // 22:21

Quote:

Originally Posted by deluxe

I'm not impressed with performance in current days games. My Q6600 @ 3.8 still outperforms these new cpu's.
Not worth the money yet to start upgrading now. DDR3 currently also isnt much better than the latest low latency DDR2.

gaming performance

Core i7's performance in current games isn't much better than Core 2 Quad. Games cannot utilise all the cores and hyperthreading etc, and the rework of on die caches isn't the best for gaming. However, in some next generation games/engines, namely Lost Planet, show impressive gains: around 26% (Core i7 3.2Ghz compared to Core 2 Quad 3.2Ghz). I agree with you though: gaming performance isn't impressive.

Also DDR3 is a lot better than the best DDR2: the memory bandwith offered by triple channel DDR3-2000 is amazing.

Quote:

Originally Posted by Admael

I assume the "new overclocking methods" is referring to this?

http://www.theinquirer.net/gb/inquir...damage-nehalem

No, not at all. The "new overclocking methods" refers to the actual methods to overclock the processors. Read this article if interested. And that Inquirer article, like any of the articles on Core i7 memory controller voltage issue, has been outdated considering Core i7 has been released.

Admael · Nov 05, 2008, 07:55 AM // 07:55

I was being facetious, and that article is outdated, yes, only because the work-around is simply, to "run the RAM at manufacturer's listed rates and voltages", which defeats the purpose of overclocking!

Quaker · Nov 05, 2008, 04:31 PM // 16:31

Quote:

Originally Posted by Cyb3r

Little effect on gaming quaker ya seen the Scores in the guru3d article with Sli compared to the old ones?

I know I shouldn't even bother to point out the obvious, but I will.

I was referring to the i7 965 vs QX9700 (Core 2) in the table at the start of the thread.

Lord Sojar · Nov 08, 2008, 08:08 AM // 08:08

So I had the thrilling opportunity to try out a Nehalem test bed on the eVGA x58 motherboard along with an unnamed GPU (which shall remain unnamed) a little over a week ago.

I must say, Nehalem is very impressive. I had few issues, except in the area of overclocking. Now, that is not to say overclocking is anymore difficult on Nehalem; rather, it is simply different and a bit more time consuming.

Nehalem really doesn't like anything above 1.7v on the DIMM interconnect... it starts to act strangely, in ways I have never seen a processor act up. Cache misses seemed to be the biggest issue. Nehalem, however, runs very cool. It does appear Intel was able to really perfect their 45nm fabs, and streamlined them for Nehalem.

In the category of gaming, it is hit or miss. In many modern games (that I actually got to run with the unnamed GPU), Nehalem seems to thrive and best its predecessor. However, in games 1 year+ in age, Nehalem is equal or slightly under the performance of Penryn. I am sure the reasoning behind this is, in fact, the different memory hierarchy, and how Nehalem handles cache.

Speaking of caching, Nehalem is very very odd on that topic. It doesn't seem to have any particular pattern when it comes to caching data. It seems to have a mind of its own when it decides which medium it is going to cache data to, and which command sets it wants to utilize. I don't think this effects performance, but I did find it odd to say the least. The branch prediction and OoOE implementation was phenomenal, even improved beyond that of Penryn. Matrix calculations were superb, probably the biggest improvement as far as large scale calculations are concerned. Nehalem is phenomenal on database work, as well as macro scale simulation.

In summary, from what I could tell, Nehalem offers some very nice performance boosts. However, from a pure gaming standpoint, it may not be worth the price to upgrade at this point in time. Q1 2009 may very well be a better time to throw the money at a sizable upgrade. If you have any questions, feel free to ask them here. I will do my best to provide an answer, provided I had time to experience said question or something closely related in my short time with Nehalem.

Brianna · Nov 19, 2008, 08:35 PM // 20:35

http://www.evga.com/articles/00438/

Stuff is out on www.newegg.com now, $1,000 USD for the Extreme, but I'd expect that.

bhavv · Nov 20, 2008, 09:18 AM // 09:18

Good to know tht there isnt hardly any difference over core 2 duo and dual channel memory in real world applications and gaming.

I was initially interested in Nehalem, but after seeing the reviews, I'll skip it and wait for the 32 nm die shrink before upgrading again.

The oblivion result of no multi GPU scaling on X58 is rather off putting too.

Brianna · Nov 20, 2008, 09:36 AM // 09:36

Yeah the 32nm is a lot more interesting to wait for.

Whats the no-multi-gpu scaling about though? (Not exactly keeping up with all this).

Kattar · Nov 20, 2008, 12:15 PM // 12:15

Quote:

Originally Posted by Rahja

Q1 2009 may very well be a better time to throw the money at a sizable upgrade.

Dates, Rahja, dates. I need them. Specific ones too.

Snograt · Nov 21, 2008, 11:25 AM // 11:25

EVGA's X58 boards look rather nice
http://www.evga.com/articles/00438/
Tried their E-LEET tuning doohickey, Rahj?

Evil Genius · Nov 22, 2008, 02:51 AM // 02:51

Quote:

Originally Posted by bhavv

The oblivion result of no multi GPU scaling on X58 is rather off putting too.

You have got to be joking!

Complaining about the performance of multi GPUs and a new CPU microarchitecture for a 2 and a half year old game!

Why is this the stupidest comment I have seen in relation to Core i7, on any forum, ever?

1) A single modern GPU such as a 4850, 4870, 9800GTX, GTX 260 would have no problems running that game at any resolution and details settings (hmm maybe the 4850 and 9800GTX would struggle with Oblivion at 2560*1600 4AA 16AF). A single 4870 or GTX 260, coupled with a simple Core 2 or Core i7, would tear oblivion apart at any res and detail settings. Why would you need multi GPUs for Oblivion?

2) Its the game rather than the hardware. Why would a game over two years old benefit from the SMT of Core i7? The 4 cores? The DDR3 memory?

3) Core i7 dramatically improves multi GPU performance in many modern games. To complain just about Oblivion...gees some people will never be happy.

Quote:

Originally Posted by Guru3D

Pun aside, the 3.2 GHz Core i7 processor used in this article seems to be a very good match for heavy weight Multi-GPU environments, whatever your preference is: 3-way SLI or QuadFire. You will not likely run into CPU bottlenecks anytime soon. That's of course until ATI and NVIDIA release faster cards again, which is bound to happen anyway. But in retrospect the Core i7 platform is a fantastic platform for multi-GPU gaming as you'll gain heaps... seriously heaps and heaps more performance opposed to today's regular PCs with a Core 2 Duo processor.

Kokuyougan · Nov 22, 2008, 03:31 AM // 03:31

I think he didn't mean the game Oblivion, but oblivion as in bad or horrible. :x

He's wrong, either way. I think.

So, it's a lot better if you have...say 260revB or 4870 in SLI/CFX and a 920/940/965 [all of them, or just the extreme?] than if you had a core 2 duo/quad and the same in SLI/CFX? Is it really worth waiting for the 32 die shrink? WTH does a die shrink refresh do anyway?

Evil Genius · Nov 22, 2008, 04:06 AM // 04:06

Quote:

Originally Posted by Kokuyougan

So, it's a lot better if you have...say 260revB or 4870 in SLI/CFX and a 920/940/965 [all of them, or just the extreme?] than if you had a core 2 duo/quad and the same in SLI/CFX? Is it really worth waiting for the 32 die shrink? WTH does a die shrink refresh do anyway?

Read the Guru3D article that compares an E8400/Quad 9770/Core i7 965 and 260/4870 etc. Link. What applies to the 965 would apply to the 920 as well, albeit to a lesser degree.

Not worth waiting for the die shrink in my opinion (its about a year a way at least after all). Die shrinks mean: lower voltage/power consumption, better overclocking and larger cache.