And both of those reviews are based on the older drivers. Also, the config used sub 4ghz processors, which can impact the results in a big way because Nvidia hardware is more dependent on CPU power. The HardOCP redux test proved that.
You shouldn't really test a 580 SLi rig with any CPU below a Core i7 @ 4ghz imo (but preferably 4.2ghz and above), due to CPU bottleneck.
I found this statement interesting and thought I would check it out for myself. I duplicated the test settings used here at Anand except my CPU is clocked to 4.2Ghz, while their test has the CPU at 3.33Ghz. I used two 480s clocked to 850/2000, which is a bit faster than two 580s.
Anand's test used the 262.99 launch drivers, I am using the latest drivers; 285.38.
I show 6fps more on minimum and maximum vs Anand's results. That is a year of new drivers, cards a little faster than stock GTX 580s and a CPU nearly 1GhZ faster. Not nearly as significant as you seem to think.
I was actually curious what sort of gains drivers may have brought to Tri-SLI as well, so I compared my system to Anand's GTX 580 Tri-SLI results.
Anand's 266.58 GTX 580 Tri-SLI benches:
My results using the most recent 285.38, I'm actually slightly slower here. So some performance was lost in Crysis with the most recent drivers vs these older ones.
This is why I don't even trust most review sites these days, and prefer to get first hand opinions from gamers that actually own the hardware.
Here's an example from a guy that does own both a 6970 CF and 580 SLi set up
You've discounted reviews from Anandtech, Hardocp, Guru3d and Driverheaven in favour of an anecdotal post from a random forum user. I find this a confusing way to form a solid opinion.
Anand is easily the most unbiased and best review site on the web. I put much more stock in their results than a random forum post, especially when I see the results mirrored by other review sites.
A lot of people that go with AMD initially for the price/performance advantage regret their decision, and then switch over to Nvidia.
Was this your experience having used AMD cards and why you switched to 580s ?
In the case of what we are discussing here with AMD vs Nvidia at 2560x1600, it's not even a true price vs performance discussion. They perform equally at this resolution but AMD is much cheaper. It's just a price discussion. :thumbsup:
I used 5870s before my 480s, they worked flawlessly and my experience with AMD's drivers was better than my recent experience with Nvidia's drivers. I've been having trouble with poor multi-gpu scaling and poor performance in several games; having to wait, sometimes as long as two months for fixes.
In single card config perhaps, but not multi GPU. Nvidia are always tweaking their SLi profiles.
That may be. But Crossfire definitely has the edge over SLI these days. Just look at the 6950 CF performing as well as 580SLI. The CF scaling is much better.
Crossfire scaling is 80% in Crysis:
http://www.anandtech.com/bench/Product/293?vs=299&i=188.191
SLI Scaling is 65% in Crysis:
http://www.anandtech.com/bench/Product/308?vs=305&i=188.191
It is sort of surprising how well AMD does in Crossfire with 6970s vs 580s in SLI at 2560x1600, 5760x1080 and other high resolutions. Especially with how much cheaper the 6970s are vs the 580s.
It's a mix of better texture throughput, more memory and better crossfire scaling. If the 6970s had been available when I upgraded to 480s, I would of gone for them over my current cards in a heartbeat. Unfortunately they weren't and there has been nothing worth upgrading to, so I am waiting on the new 28nm GPUs.
For someone shopping for a multi-gpu 2560x1600 setup, 6970 CF is the one to get.
Facebook
Twitter