EVGA GTX 660 Ti Superclocked SLI Scaling

Let me start this article off with an apology – I was unable to complete overclocked runs of the cards in SLI, thus, I’m sorry. One of the cards needed to make its way back to NVIDIA to look at an unexpected heat-related issue (only affecting that one card as far as we can tell) and that issue, combined with very little time to bench them, ended up being an insurmountable obstacle. Things were so rushed, I didn’t even remember to take a photo of the two cards installed together. Anyway, that’s why all of the SLI results you will see here are completed at stock clocks.

So, with that said, let’s start the show. This should be pretty short, sweet and to the point, just presenting the data for you to see. The card itself you already know from our review of the EVGA GTX 660 Ti Superclocked; this is just two of those together in SLI.

As you can see, they make a heck of a pair for 3DMark 11, which is right where we’ll start – Synthetic Benchmarks.

Synthetic Benchmarks

When going through the benchmarks and games today, we’ll just present the data in a constant stream, then stop to check out how things scaled.

3DMark03 wasn’t much of a surprise, I didn’t expect that to scale all that well. Vantage was quite the surprise though. DirectX10 isn’t that old, however, remember that Vantage is a CPU-bound benchmark on newer platforms. The single-card and SLI GPU scores from Vantage results were 29713 and 48200 marks, respectively; so the actual GPU part of the bench scaled very close to where 3DMark 11 did – 162.2%.

Now, when you get to the DirectX 11 benchmarks, which are still pretty solidly GPU-bound, you start to see a good separation. Heaven was quite impressive, almost fully doubling the score.

Game Tests

For our game tests, we’ll split them up into two sets of three so you don’t have to look at a massive wall of graphs.

Well then, these first three games show massive scaling, with darn near double framerates for AvP & Battlefield 3.

There, wasn’t that one sentence to break up the wall of graphs nice?

Not bad at all for these three but not as good as the first three.

As you can see in the scaling chart, the curve is pulled down quite a bit by Civilization V. If you removed Civ V, which doesn’t scale well at all, you get a much more respectable scaling average of 182.2%.

Now let’s hook up a couple more monitors and see how Surround scales.

NVIDIA Surround

The GTX 660 Ti is a little bit more limited here because of memory bandwidth, but even so you can see strong scaling in pretty much everything but Metro 2033. Even with a pair of 660 Ti’s, Metro brings the system to its knees.

Still, with Metro’s pitiful scaling, you get a strong scaling average. Civ V even scales well when run at such a high resolution. Again removing the worst performer (Metro 2033), you get 179.4% scaling.

Thus ends our exploration of the EVGA GTX 660 Ti SC times two. NVIDIA has done well with this GPU’s SLI scaling. With the rare exception, which depends on resolution, you can expect very near 180-185% scaling throughout.

At $600+ for a pair of these, if you have that much money to begin with, I think it would probably be a better decision to go with a GTX 680 (which start at $480) or HD 7970 (which start at $400). If you read our recent review of the ASUS HD 7970 DirectCU II TOP ($440 MSRP) and saw how well it performs in Eyefinity (its overclocked framerates were overall north of the pair of GTX 660Ti’s), you can see why I say that.

However, if you only have $300 and need a gaming system right now, grabbing an EVGA GTX 660 Ti SC for that much now and upgrading it later by adding a second GTX 660 Ti wouldn’t be a bad one-two punch in the fight for higher FPS.

–Jeremy Vaughan (hokiealumnus)