SLI vs Crossfire on Core i7 975
IF WE PUT ASIDE the dual-GPU monster cards, the other way to get parallel graphics on your PC is by combining two single GPU cards. Now, finally, we have a way to compare Nvidia SLI and AMD Crossfire approaches on one same chipset, Intel's X58 for Core i7 processors.
Does this make for an easier apples-to-apples comparison? Yes, since in prior situations you also had to take into account chipset feature and performance differences.
Right now, the top single GPU cards in each camp are GeForce GTX285 and Radeon HD4890. Both have 1GB video memory as standard, even though there are 2GB versions on offer, and teraflop-class single precision FP. I put them together running on a Core i7 975 3.33GHz setup with an Asus Rampage II Extreme mainboard and 6GB of Kingston HyperX DDR3-2000 Core i7-optimised memory.
Windows Vista 64 and new drivers, ATI Catalyst 9.5 and Detonator 185 were used for the 3Dmark Vantage runs. Each run was done in both SXGA 1280x1024 Performance mode and WUXGA 1920x1200 Extreme mode, in dual card setups. Here are the results:
Now, notice something, to avoid results skewing due to the PhysX, I ran a series on Nvidia with GPU PhysX disabled as well.
In summary, even with PhysX disabled, Nvidia still has a slight edge in total performance. Do keep in mind, though, that the ATI cards are cheaper right now. Simply, ATI's manufacturing costs are lower, and a 256-bit memory bus is somewhat easier to implement than 512-bit wide memory. So, at least if you look at 3Dmark, you get what you pay for, that is, Nvidia is more expensive, but it's proportionally faster too.
Long-term reliability, on the other hand, might be another matter. µ
Share this:
ShareBusiness Intelligence Software
Customer Relationship Management (CRM)
Enterprise Accounting Software
Enterprise Resource Planning (ERP)
Enterprise Systems Management
I do love the nVidia bashing :-)
A bit stupid to take Vantage as performance meter. We all know that nvidia has advantage there. Put real world games to test and suddenly there isn't much difference.
I was wondering about the following points..
What is the most efficient set up of the two. In other words.. which of the two setup comes is scaling closest to a 200% score over a single card set up?
Maybe a few games in the mix?
Maybe a little more info about items like the relative cost and power usage..
What they dont tell you in the reviews of the latest x2 card with the latest games is the display issues with legacy games. Here I am pointing at ATI, 3870x2 generation to be precise, one of which I have had since new and will be getting rid of as soon as I can justify it.
X2 never again! They have great 3D benches in some games but dont display 2d text and some textures properly in many games that are not making the headlines in the gaming news. Disabling catalyst AI doesn't always fix it and some glitches are insoluble meaning you cant play some games properly! The horror!! With two seperate cards in Xfire you can disable one and circumvent the dual GPU debacle.
I think that is not balance if HD4870 CF vs GTX285 SLI, maybe better if the comparison between HD4870 CF vs GTX275 SLI. Because 4870 have the same price with GTX275 and they performance of single card is nearly same.
This seems like as good a forum to bring this up as any-
I thought the reliability issues with the nvidia GPUs were confined to a select group that didn't include mine (please correct me if I'm wrong). I've got a Dell XPS 1730 with dual 8800M GTXs. I've owned it for JUST over 12 months and one or both of the cards just shat itself a couple of days ago. Hugely impressed.
Thankfully Dell has said they're sending a tech out to replace them this week. My only issue is that if they last for another 12 months and then die, that'll be just outside my warranty. It's going to be a while before I buy an nvidia card by choice for years to come. They've got some trust to build back. I'm sure I'm not the only one.
I just bought a 4890 for my desktop, so hopefully I won't be let down there.
Not too bad a test..
Not exactly the Be All and End All test though.
On average a 2x 4890 setup is at least between USD $100 to $130 cheaper than a comparable 2x GTX285 (no i didnt consult the latest prices on the egg, just prices i have seen from retaliers over the past few weeks).
The synthetic benchmarks usually favour one maker's cards/ architecture. this isnt because the were paid money, just how the software was written, taking in the advantages of certain functions within both architectures.
remember the 3870X2. It was a relatively shitty card compared to nvidia offerings in terms of fps in games. but it was the bencher's choice of grahpics card when it came to 3D mark 06.
Similarly the Vantage Benchmark is an Nvidia dominated benchmark. Benchers will choose the GTX series over the 48xx series.
Im not being biased to either company. But synthetic bencharks do not reflect real term performance in games (which is what people play, unless you are a Tier 1 bencher using LN2 like i am and have your cards clocked at a minimum of 1ghz core).
For games i would pick a 2x 4890 setup over a 2x GTX285 setup. because on average a GTX285 is only 10% faster than a 850mhz core'd 4890. granted you can overclock both company's cards, if you raise the core speed of a 4890 to 950mhz (which is easy enough) you will surpass GTX285 performance. With current prices set to drop by up to USD $50 per card and with rebates as ati clears their inventory for their DX11 arrival it makes the offering even more alluring. saving up USD 200 over the GTX setup.
however if you bench 3d mark 01 or Vantage like me, GTX285 Tri sli or GTX295 at 1ghz core is the way to go!
But i wouldnt go all out with a GTX200 series or 48xx series setup.
GT300 WILL be available at Q4 2009. My sources tell me that Nvidia is adamant about this Q4 launch. Even if this means a slight paper/low retail availability launch. This could be as late as late december. But its still Q4 :P.
It wont be a g80-like part, but have money on that it will beat ati offering by a larger percentage than this current round. However, dont expect them to be cheap! ;)
:D NDA is useless if you have anonymity :D
sorry for my bad english
Does anyone think AMD is faster than NVidia? I mean seriously? In real world gaming bench marks NVidia consistantly pulls out slightly better numbers. Thus to point out NVidia as faster is complete foolishness.
Value. Where does one card shine over another. If you mostly game and have a budget of X, perhaps AMD is the best option? There's what a $200 difference between these two rigs.
How about some in-depth look at spacific applications which use each card? Blu-Ray acc? Photoshop acc?
Thanks for doing a poor job telling us what we already knew...
There are Two Models of FireGL v3700. OLD one with part number ending in 51 & NEW One, with Open GL 3.0. Its Part Number Ends in 62. Both Cost Same. About $100. Pci-e 2 shader 4.1 Dx10.1 & list gos on. 3D Rendering on Economical Workstation. Todays Consumer Game Cards SUCK. They are over Specificationed, When IN Fact, Often won't even Play Vantage NOR Even Simplist of Games, Let Alone Complex Monsters.
Get with IT. WHY RIP READER OFF, WHEN EQUALLY COSTLY CARD IS MADE BY SAME COMPANIES.
ATI FIRE GL v 3700 that retailler with egger in its name has 'em by 99case for $74 in orders of 6. FIRST LINE; BRAND NEW.
The only thing Vantage is good for is a survival test for your new hardwares.Otherwise, this software sucks when it comes to actual performance comparisons.
Vantage is heavily optimized for nvidia cards.And it shouldn't be used to make a performance comparison in any way.
To Highlander.
The 4890 is on par with the gtx285 at 1920 x 1200 resolution and below.At 2560 res, the card falls behind due to its smaller memory interface which becomes the limiting factor when dealing with the big textures at that resolution.
Here, check this review from this German site:
[urlhttp://www.pc-experience.de/wbb2/thread.php?threadid=28986[/url]
or check this review from xbitlabs:
[url]http://www.xbitlabs.com/articles/video/display/radeon-hd4890.html[/url]
umm... the SLI setup pretty much destroys the crossfire rig.
Vantage is a useless test.It doesn't even support crossfire properly.Here's a video from my pal of 2 Hd4890's Crossfire playing Crysis Paradise Lost (snow level) Max 8xAA.check it out.
http://www.youtube.com/watch?v=vmGKSCVW2uU&feature=related
The guy is averaging around 45fps.
Show me 2 gtx285s doing that.
Only synthetic benchmark geeks care about synthetic benchmarks. Show us some FPS gaming scores and I'll be interested. As for long term reliability, only time will tell.
this is all bullshit.. cause nVidia might be better in its class..
but if nvida costs alot more then ATi would expect them to be better..
since u can get a TWO 4870X2s for the price of one 295GTX.. ATI is better performance.. since u can buy more, for less.
wat gives ?
I dont really understand the chart, in terms of cpu its losing like HALF , but in terms of GPU its like 35% ,
why the hell cpu score is higher on nvidia? I tot both using same platform to test?
While ATi is cheaper than nVidia, they also have a habit of sucking out much more power. So I think in the long term, ATi will end up costing more on the electrical bill than nVidia will.