Core i7: 4-Way CrossFire, 3-way SLI, Paradise?

[cangelini]

[citation][nom]Tiger King[/nom]very nice article.but upon reading article from another site, i'm wondering if the editor actually used the latest engineering ATI driver.check out Guru3D's article:in that article, ATI and Nvidia cards seem to perform much closer to each other in terms of performance. and the editor on Guru3D said that there was hardly any kinks during the testing.if anything i've concluded is wrong, please correct me, as i'm a very casual computer person[/citation]

No worries--yes, we used the latest driver. In fact, we went through three separate drivers in the time the story was started until it was finished, retesting along the way. AMD knows about the issues we've encountered and we're still waiting for additional input from the company. Thanks for the questions!

 

[Guest]

Yea...I have Company of Heroes running at like 121fps with 2x 4850's...and X2 6000+ cpu.....so something very wrong with your results in that department....just fyi.

 

[randomizer]

[citation][nom]Major Error[/nom]Yea...I have Company of Heroes running at like 121fps with 2x 4850's...and X2 6000+ cpu.....so something very wrong with your results in that department....just fyi.[/citation]
Maybe you used drivers that worked, or you weren't using the ingame benchmark to measure average framerate which is what I assume was used in these tests.

 

[kelfen]

expensive like all new things are but looks promising!

 

[agawtrip]

that's it?!?! for i7!?!?

 

[MattC]

I know this probably isn't the best place to say this, but since two others (or more?) have already pointed it out here, I have to agree that the scroll bar used to select the various pages within an article is horrible. I have hated it since you guys adopted it - honestly I'd rather have a list of hyperlinks at the bottom. Yes, it takes up more real-estate, but it's simple and it works, and who cares if a long page gets 20 lines longer? More space for ads on the side, right?

 

[cangelini]

[citation][nom]mattc[/nom]I know this probably isn't the best place to say this, but since two others (or more?) have already pointed it out here, I have to agree that the scroll bar used to select the various pages within an article is horrible. I have hated it since you guys adopted it - honestly I'd rather have a list of hyperlinks at the bottom. Yes, it takes up more real-estate, but it's simple and it works, and who cares if a long page gets 20 lines longer? More space for ads on the side, right?[/citation]

Oh, you guys are talking about the drop-down. I'll jump on the bandwagon and say I hate that as well. One of the interface "features" that is known to be disliked and needs to be changed. I'll pass the table of contents idea along Matt, but I know that is one of the options already being discussed.

 

[randomizer]

I don't find the dropdown much of a pain, it works for me. It just requires the page to completely load before it can be used, which on this far away island under the sun, is quite a while.

*Requires Australian localisation*

 

[bf2gameplaya]

I can not shake the fact that when I look that Tom's main page graphic for this article: It reads due to the font, letter placement and slant:

Core i7 Gaming: It's Broken.

 

[zaeentech]

The review without the Flight Simulator X (FSX) Benchmarks is very much unexpected from Tom's Hardware and isn't complete. We the flightsim fans are eagerly waiting for the FSX Benchmark from you. So Guys please update this review by including the FSX Benchmarks data under Intel Core i7 965.

 

[anartik]

[citation][nom]randomizer[/nom]I don't find the dropdown much of a pain, it works for me. It just requires the page to completely load before it can be used, which on this far away island under the sun, is quite a while.*Requires Australian localisation*[/citation]

I would have to agree... It only works when the page fully loads which is easier said than done. Since back to probably the last remake of Tom's there have been serious functional and performance issues with the site. Most of it seems to be related to external links to ads. Very often I have to hit the stop button to even be able to read the page and then the pull down menu does not work or the comments are totally garbled. When a page does fully load it is very, very slow most of the time.

 

[mapesdhs]

I find the page loading of advert data makes toms very slow now,
and often I have to click Stop, or the browser reports a hung
script anyway. Other sites load much faster.

Ironically though, I don't see this scrollbar thing at all as
I'm using Firefox on an SGI.

I just click Stop once I'm sure the main text has loaded,
otherwise the ad loading prevents the page from being scrolled
for ages.

Ian.

 

[Guest]

Hi all...is it me or after reading this article, if we look at page 13 ("Averaging It All Out"), or almost every individual benchmark, we see that having 1 GTX-280 performs better than having 1 4870x2('2 x Radeon HD 4870') ? I know that drivers will change (hopefully on ATI side...), I know it's a bit early to base our purchase on this review since the platform is brand new and things will probably change in the next few weeks, but based on this article, 1 GTX280 seems better than 1 4870x2, for the X58 platform and for the games being tested...anyone to convince me that I'm wrong with this conclusion...?!

 

[mapesdhs]

frankydude,

I examined the data for you, here are the results, ie. i7-965
platform with 4870x2 vs. single GTX280, in each case for each
resolution and mode, answering the question, "Which is faster? The
GTX280 or 4870x2?"

World in Conflict, 1920x1200, AA/AF OFF: 4870x2
World in Conflict, 2560x1600, AA/AF OFF: 4870x2
World in Conflict, 1920x1200, AA/AF ON: 4870x2
World in Conflict, 2560x1600, AA/AF ON: 4870x2

Supreme Commander, 1920x1200, AA OFF: 4870x2
Supreme Commander, 2560x1600, AA OFF: 4870x2
Supreme Commander, 1920x1200, AA ON: 4870x2
Supreme Commander, 2560x1600, AA ON: 4870x2

Crysis 64-bit, 1920x1200, AA OFF: 4870x2
Crysis 64-bit, 2560x1600, AA OFF: 4870x2
Crysis 64-bit, 1920x1200, AA ON: 4870x2
Crysis 64-bit, 2560x1600, AA ON: 4870x2

Crysis Warhead, 1920x1200, AA OFF: 4870x2
Crysis Warhead, 2560x1600, AA OFF: 4870x2
Crysis Warhead, 1920x1200, AA ON: 4870x2
Crysis Warhead, 2560x1600, AA ON: GTX280

Company of Heroes, 1920x1200, AA OFF: GTX280 (ATI driver problem)
Company of Heroes, 2560x1600, AA OFF: GTX280 (ATI driver problem)
Company of Heroes, 1920x1200, AA ON: GTX280 (ATI driver problem)
Company of Heroes, 2560x1600, AA ON: 4870x2 (ATI driver problem)

Unreal Tournament 3, 1920x1200, AA/AF OFF: 4870x2
Unreal Tournament 3, 2560x1600, AA/AF OFF: 4870x2
Unreal Tournament 3, 1920x1200, AA/AF ON: 4870x2
Unreal Tournament 3, 2560x1600, AA/AF ON: 4870x2

Far Cry 2, 1920x1200, AA OFF: GTX280 (ATI driver problem)
Far Cry 2, 2560x1600, AA OFF: GTX280 (ATI driver problem)
Far Cry 2, 1920x1200, AA ON: 4870x2 (ATI driver problem)
Far Cry 2, 2560x1600, AA ON: 4870x2 (ATI driver problem)

3DMark Vantage, 3DMark: 4870x2
3DMark Vantage, GPU: 4870x2

Average Scores, 1920x1200, AA/AF OFF: GTX280
Average Scores, 2560x1600, AA/AF OFF: 4870x2
Average Scores, 1920x1200, AA/AF ON: 4870x2
Average Scores, 2560x1600, AA/AF ON: 4870x2


I hate to say it but I can only conclude you've been misinterpreting
the graphs. A 4870x2 beats the GTX280 80% of the time (ie. 24 out of
the 30 comparisons), and where it doesn't it's because of a driver
issue. If ATI fixes ther drivers, I'm sure it would win in virtually
every case for this specific question.


Note that the GTX280 has a multi-GPU driver problem for Crysis 64bit,
but your question wasn't about SLI performance so that doesn't metter
for this analysis. ATI has problems with scaling in Company of Heroes,
and in Far Cry 2 when AA is not used or the resolution is high. In time
these problems will presumably be fixed in all cases.

Of course, the results would be different if comparing 2 x GTX280
to 4870x2, but that wasn't your query.

Hope this helps!

Ian.

 

[mapesdhs]

Oh, I didn't include the 3DMark Vantage CPU numbers because the
use of PhysX gives NVIDIA an incomparable advantage.

 

[tj_the_first]

Interesting in a theoretical way. I couldn't afford that Intel CPU/Mobo until 2010 and by then who knows what graphics there'll be. Extreme articles like this shouldn't really influence the average gamer.

In Australia the QX9770 sells for over $2000 and you can buy an entire Phenom + 4870X2 system with that money.

 

[mapesdhs]

Likewise, in the UK the QX9770 is over 1000 UKP, whereas, despite
much a good dealer slower, the Phenom 9950 Black Edition is only
140 UKP, though in reality one would probably at this point go
for the Q9550, E8x00 or Q6600. The Extremes are real money
burners. So far it looks like the performance of the i7-920 means
that the older generation of Extremes will have little relevance
for many tasks, though an older build may still be more logical
wrt price/performance if we don't see reasonably priced X58 mbds
and affordable DDR3, ie. despite the speed disadvantage in some
situations, consider the overall low cost of an E8400/Q6600 with
4GB DDR2/800 and reasonable mbd compared to initial X58 offerings.
In other words, it looks quite possible that for the price of a
single i7-920 or 940 system, one could have two E8400/Q6600
systems, both fitted with decent coolers for overclocking. Not
so relevant for games perhaps, but my main task is video encoding
so 2 systems for the price of 1 is a very appealing option. i7
has the speed, but will the overall system cost be attractive?
Right now decent DDR3 looks to be 3X to 4X the cost of DDR2,
and will there be X58 mbds in the $100 to $200 price range?

Ian.

 

[mapesdhs]

Brief followup...

Decided to do a search to check, best price I could find for
the QX9770 in the UK was 917 UKP:

http://www.evilgaming.co.uk/product.php/intel_core_2_extreme_qx9770___3.2_ghz_processor

Still too high, but of course there are those who are happy to
pay the premium on these unlocked parts.

Ian.

 

[maxinexus]

my 2 superclocked 1GB 8800gt in SLI for total of $240 work flawlessly on 1920x1200...why spend $$$$ if $$ gives you what you need

 

[falchard]

Hooray for AMDs awesome performance despite people depicting it as a dead horse. This benchmarking shows how AMD fairs on thier peak system against Intel's peak systems on similar parts, that being the AMD 4870.

If we look at the averages. AMD performs about 16.6% worse then the Core i7 965 Extreme Edition, and 16.7% worse then the QX9770. At a higher resolution, this gap becomes smaller with AMD at 12.9% and 12.5% worse respectively. Its also closer with fewer video cards as it gets 8.5% worse at the highest resolution with the core i7, and 6.1% worse then the core2.

This performance difference comes at a price tag $985 cheaper then the Core i7's projected release value, and $1345 cheaper then the QX9770. When you factor in the processor and memory since they are the 2 only major differences between the systems.

I think it shows when money is a factor, there is room for a Phenom 9950.

 

[mapesdhs]

As always, there will be those who buy based on price, those based
on performance, while most users just wade through the mountain
of options by considering both sides of the equation. Few sites
bother mentioning price performance when reviewing CPU/gfx
products, which is a pity.

Spending more gives one more speed, but the degrees of improvement
tail off rapidly. Beyond a certain point, most would not bother,
but there will always be a market for those who are happy to
spend enormous amounts for the perceived maximum possible speed
(emphasis on 'perceived' btw; I'd be surprised if the big budget
buyers always make the best decisions).

maxinexus, you're right about the speed gains from lower cost
SLI, though not everyone wants to go down that road, or is able
to. Bit of a pain that drivers are such an issue for multi-GPU
setups aswell, seems that SLI/Crossfire performance is often
very application specific, good for one game and bad for another.
Neither technology has yet become a reliable generic parallel gfx
solution. Really, it's absolutely outrageous that one could in
theory spend $6000 on a mega-system, aiming for insane gaming
quality, only to find rubbish performance because of driver
problems for the particular game one is playing. Funny how
computing is the one area of consumer technology where the
software that drives a system component is _expected_ to have
numerous flaws on initial release. Nobody would tolerate such
a high degree of problems in a car or hifi unit.

Driver development costs money though. Interesting thought:
would you all be happy to pay, say, 20% more in order to have
more reliable and bug-free drivers on first release of a product?

Ian.

 

[anartik]

"Driver development costs money though. Interesting thought:
would you all be happy to pay, say, 20% more in order to have
more reliable and bug-free drivers on first release of a product?"

LOL I have been building PC's since the mid 80's and the the problems have never changed. Manufacturers, particularly in certain less than ethical societies, only care about getting the product to market taking the easy path of we'll fix it later (read as total lack of QA). This trend has only gotten worse with the availibility of high speed internet. I think damn near every componant of my last build required BIOS/firmware/software updates to iron out the kinks.

Lot so long ago a typical good motherboard and graphics card ran maybe $100 each. Now we pay 3-5 times as much and still get the same crap. There are certain countries that are pretty good-excellent at building hardware but not a one can write software to save their ass. I think they live by the mantra "if it compiles ship it".

 

[mapesdhs]

I was told by an admin guy at a movie studio that this is why
pro cards like the NVIDIA Quadro FX5600 cost as much as they do,
ie. much more driver development as opposed to any significant
hardware differences to the mainstream cards, except for internal
optimisations, RAM size and scalability features.

You're right though, it does seem like NVIDIA doesn't really
bother trying to offer a quality card in the sub-$200 range
anymore. The GTX is great, but far too expensive, while the
8600GT was crippled, just not worth the lower cost compared
to the 8800GT. Why isn't there a GTX at a decent price point?
8800GT SLI still looks better than a GTX260. Meanwhile, the
persistant muddle of previous NVIDIA cards is crazy.

Hmm, who knows, perhaps the economic climate will result in
NVIDIA having to lower its pricing bands next year, at least
for the midrange products anyway.

Ian.

 

[mapesdhs]

Extra thought: same applies to decent coolers on gfx cards. So
yeah, we buy on price but get a card that runs hot as hell. When
I bought my first decent card (X1950Pro AGP) the temps were sky
high. If there was an alternative version available with the
ACCELERO X2 cooler already fitted, I would definitely have bought
it for sure, given the incredible difference it made (load temps
down by 35C). The cost of the better cooler wasn't very much,
and I'm sure review articles would've conveyed the added value of
getting a card with a good cooler already fitted (which is why
I went for the Gigabyte 8800GT GV-NX88T512HP when upgrading as
it comes prefitted with a much better cooler, even though it was
not quite as fast as other stock-cooled oc'd cards that had no
headroom for oc'ing such as the ASUS TOP).

Seems like a lot of this stuff is unnecessary kerfuffle designed
to make it seem like we're getting a good deal when in the end
we have to buy extra bits anyway. Ditto the world of stock CPU
coolers, though the Nehalem's default cooler seems to work quite
well. By contrast, the cooler that came with the 6000+ was terrible.
Xigmatek has shown that good air cooling solutions don't require
one to strike oil to afford them.

Ian.

 

[geckoar]

Well all I know is that my Phenom 9950 oced to 2.85ghz and a 512mb 4870 Can play all of my game Maxed out 8x AA/ 16xAF at 1080p res. And still have over 30fps...All games but crysis warhead...high setting with 2x AA.

So I dont need A Core i7 Because my 9950 does a great job for me. And for a lot less money.