Well, it's true. I know where he's coming from though, and actually, I completely agree with him. Titan is the better card most of the time, unless you get a good 780 vs a bad Titan. But that's basically what I am saying. When it comes down to a few mhz, you really have to decide if that is worth the premium, like any other flagship product. So...I agree with everything Alatar has to say about it. I just don't think that that premium for the subjectively small performance increase is worth it, considering my financial situation. Someone else might think it is worth it, but it isn't to me. I would never argue that the 780 is all-around better, or even the better performer half the time. But it's a stretch to recommend it to anyone on less than a UHD surround setup at the moment. Back when there was no cheaper option, I jumped on a pair of Titans. Now that the available options have changed, so has my opinion. But Titan really is still the better card if you don't factor prices in.
Well, here's my little mini-review, offering the few opinions that I have and the benches I had time to run. The 780 numbers are from the ones I have now, the EVGA ACX's, and the 680 Lightning and Titan numbers are from the comparison I did when I first got my Titan. They are on the Titan's launch drivers, just like the 780 numbers are (320.18). I know slight differences could be attributed to newer drivers, since the 680 and Titan numbers are from Titan's launch drivers, but oh well. This is all I have, so, guess it's better than nothing.
And first off, before any criticism comes my way...I know: my numbers are not a definitive "x card is way better than y" and are not meant to be. These numbers and opinions are based only from cards of mine, that hit these particular clocks, and I know that in a different situation, with different attainable overclocks, the results may be different. Titan's memory wasn't overclocked at the time because I was doing scaling percentages comparable to the performance gain on the core clock so that's the only thing I overclocked, and so I wouldn't have to run each test an additional time I just used those numbers for the graphs as well. Feel free to add another 3-4% onto each Titan score to account for the extra few-hundred mhz I might've been able to put on it.
Before the numbers, just want to say that these cards are my new best friends
The cooler is amazing. Even at 100% it's quiet as hell. Really soft hum too, so even though audible, it's not intrusive in the least. And this thing gives the TFIV a run for it's money, as these things stay not only quiet, but cool too. It was warm out the day I did some of the testing, and I didn't have my AC on the whole time, so don't take every temp reading as gospel. But just know that it's more than worth the extra $10 for these over the reference model.
Also...Nvidia did something
with the horrible throttling issue the Titans had on the stock BIOS, but didn't completely fix it. These 780's are MUCH better, but there will be demand for a modded BIOS still. Here's a throttle, temp in the 40C's
, Power Target well below 106%-
But, it is much better, so there's no hurry for the modded BIOSs just yet (unless they unlock a SMX, wouldn't that be cool?).
Well, like I said, I didn't exactly do as many tests as I originally set out to do. But this will give you a pretty good idea of just how close you can come to the $1,000 Titan for 65% of the price.
Here you can pretty much see an overclocked 780 match a Titan that has a pretty good stock Boost clock. This is just with my CPU at 4.6Ghz and no HT, just showing total score for the hell of it.
Here's the graphics score. This is where you'll see it match a stock Titan with a decent Boost. But of course, any Titan that isn't the worst overclocker ever will be able to pull ahead.
Metro 2033, 1080p, maxed, the works. Except for Adv. PhysX. Here, you can see the 780 sneak past the stock Titan, before it's overclocked. After being overclocked, it's a different story. But are those 3 frames really worth $350?
I'm sure the difference between the two might increase on a higher res. Maybe not though, as I don't know if Metro gets into the +3GBs VRAM usage unless you go UHD-surround. Oh well, someone with numbers from a higher res can discuss that. For now though...you can take from my review that if you're on 1080p and in the market for one card or the other, you'll be paying around $100 for each extra frame. Hard to recommend Titan to anyone that's on 1080p, maybe 1080p surround though.
Same situation. 780 surpassed stock Titan, Titan pulls ahead when overclocked. Here, each extra frame is costing you around $100 again. Whether or not that's something you're willing to pay is up to you. Not trying to argue the 780 is the better card though, as clearly it ends up losing (and again, this is one 1080p screen, and VRAM usage may increase substantially on higher res's).
Here...is quite interesting. The 780 only had roughly a 30-something mhz core clock advantage, yet was able to get by Titan by a whole frame from the extra 300mhz on the memory. This pretty much reinforces the point I made earlier: that it more often than not comes down to a small amount of mhz, depending on which has a little bit better luck with their overclock.
Here's another situation where overclocked Titan get slightly surpassed, but only by a frame. I'm certain 200 or so mhz on the Titan's memory would've been enough to take the 780, but it wouldn't have beaten it by much either way. I loved owning the Titan, it was fun, it was the only real upgrade available at the time. But I simply cannot recommend anyone on 1080p get it over the 780. Even 1080p-surround might not benefit enough to justify the premium, but again, someone else would have to run those numbers. It's starting to look more and more like the only way someone would get a Titan over a 780 today
is if they're on a Surround setup.
Miscellaneous (Click to show)
SLI is smoother than ever. I can do a wider array of games, but I just took a shot of the frametime graph during a while of Metro LL and as you can see it doesn't get any better than this. The 4 large spikes were loading screens, so they can be ignored. The rest is pretty impressive-
This is a Valley run with the awesome ACX cooler on 100% fan speed. As you can see, it's awesome
This is the big bad GK110, with 1.2v the entire run, at a max temp of only 54C. With the fan curve taking over during SLI, temps get up to 72C, but that's with the fan speed hitting a silent ~69%. Temps could be lower with a more aggressive fan curve, having to only deal with the barely audible, soft hum of this great cooler
Some 1202mhz runs, nice and stable with no throttling. Great temp on the quiet fan curve-
Metro LL with SLI. Top card levels off at 71C, fan noise more than bearable. 100fps min's with SSAA @ x2
(SLI has issues in that game apparently, usage does not stay up at 99% where it should be)-
Here you can see one of the few instances of throttle I experienced. As shown, the temp was 48C, and the Power % was at 103%. Yet, for some reason, Nvidia has this thing programmed to throttle. Needless to say, I have some questions for them, and this would be one of them-
Some SLI numbers. Didn't make a fancy graph or anything for them though, as I didn't have Titan SLI numbers to compare it to. But here's the kind of upgrade you can expect when deciding to sell your Titan and go with 2 780's instead. That's what I did, and let me tell you, $350 well spent-(also, remember, this is for an additional cost of 35% over Titan)3dMark11 Graphics Score
Overclocked Titan- 16,000
780 SLI- 27,750Increase- 73.4%Valley
Overclocked Titan- 77fps
780 SLI- 123.5fpsIncrease- 60.4%Heaven 3.0
Overclocked Titan- 78.8fps
780 SLI- 140.5fpsIncrease- 78.3%Heaven 4.0
Overclocked Titan- 61.4fps
780 SLI- 118fpsIncrease- 92%Sleeping Dogs
Overclocked Titan- 65.5fps
780 SLI- 114.2fpsIncrease- 74.4%Metro Last Light
Overclocked Titan- someone else may provide
780 SLI- 74fps (with topping out at 90% because of the issues the SLI profile apparently has)Increase- n/a%
For most people, it seems like a very worth-while upgrade, speaking from a price/performance point of view (no, I'm not speaking about the cards themselves, as I know they have a horrible p/p. I just mean the p/p when going from one Titan to two 780's. It's more than a 2/1 ratio for the most part.)Scaling- Clocks
This has the memory factored in as well, but as you can see, for the most part, scaling is much better than with GK104-3dMark11- Graphics Score
Stock to overclocked- 13,840 to 14,700Increase of +8.3% core and +5.8% memory yields 6.2% increaseValley
Stock to overclocked- 68fps to 70.5Increase of +8.3% core yields 3.7% increaseHeaven 3.0
Stock to overclocked- 70.3fps to 75.3fpsIncrease of +7.1% core and +5% memory yields 7.1% increaseHeaven 4.0
Stock to overclocked- 58.1fps to 62.5fpsIncrease of +7.1% core and +5% memory yields 7.6% increaseSleeping Dogs
Stock to overclocked- 63fps to 66.7fpsIncrease of +7.1% core and +5% memory yields 5.4% increaseScaling- SLI (is not 1000% accurate since SLI wasn't run at as high of an overclock as single-card could hit, but still relatively reliable)3dMark11- Graphics Score
SLI- 27,750Scaling- 89%Valley
SLI- 123.4fpsScaling- 75%Heaven 3.0
SLI- 140.5fpsScaling- 86.1%Heaven 4.0
SLI- 118fpsScaling- 89%Sleeping Dogs
SLI- 114.2fpsScaling- 71%Metro Last Light
SLI- 74fpsScaling- 55%
(wow, they really need to work on that SLI profile)
Well, that's what I got. Not much, but I hope it helps someone out there who's curious. Going to go finish up Last Light with these bad boys now