Krauser Kat
Member
I'll sell you a GTX 580 for $250US free shipping. no joke.
this is sooo tempting.
I'll sell you a GTX 580 for $250US free shipping. no joke.
Right. It all depends on what kind of a power envelope is acceptable. Current trends seem to be pointing at a lower TDP for Maxwell.
The 20 percent number makes no sense when you switch to a new architecture AND get a massive process node shrink like going from 40nm->20nm.
It was as big a tock as a tock can get,.
Keep in mind that Maxwell chart is performance/watts, not showing actual gains in performance. Maxwell is more than a year out. If if looks like you need to go that direction, then sell whatever cards you have and go Maxwell.I'm so torn. Should I go ahead and wait for Maxwell or take the plunge with 2 Titans? Every time I see that Maxwell chart I get pissed off because all it does is make me want to wait. I really don't know if I can wait 1 more year for my dream rig. I've been wanting to build a PC since May of last year.
Keep in mind that Maxwell chart is performance/watts, not showing actual gains in performance. Maxwell is more than a year out. If if looks like you need to go that direction, then sell whatever cards you have and go Maxwell.
Continually waiting for the next thing means that you'll never build. It hasn't been a bad time to build other than right before Sandy launch in the last... 6 years.
Okay.
So, lets say you have a single card that is rendering frames at 60 FPS. You get a second card to get FPS to 120. In terms of frame times, that is 16.7ms to 8.3ms, which is halving the time it takes to render each frame. Games poll input based on frames rendered, so that means your input is now being polled every 8.3ms instead of every 16.7ms. SLI adds in a frame of lag to help compensate for microstutter.
So essentially, you are a frame behind the action. But that frame is only 8.3ms, and it's no worse off than it was before. In fact you're still getting better overall input than you were before, as your input is being polled more frequently.
Now lets compare the dips in performance that one might see. If on that same single card, your performance falls to 25ms frame time (40 FPS). If your performance stays rougly doubled, you're looking at a frame time of 12.5ms (80 FPS). Even if there is a frame of lag, having your input polled every 25ms compared to 12.5ms is pretty unacceptable for competitive gaming. Regardless of what it is showing you on screen, the input is still being polled more frequently for the game engine to take into account.
Beyond this, you get the added benefit of having smoother action that makes it easier to track at 8.3ms on a 120Hz monitor. I'm not kidding when I say it feels like you're cheating in twitch games where tracking is important, compared to 60Hz.
Did I do the math wrong, or is Nivida going to charge $900 for performance that can be matched by a 660 ti SLI setup?
In my experience, microstutter is only a problem when the framerate does not match or exceed the refresh rate. It's also more of a problem on AMD setups than Nvidia.You can always match a high end setup for a cheaper SLI setup. A single card remains a far better solution than dual GPUs, as dual GPUs introduce a lot of problems, and microstutter can make a game feel like it has a low framerate even with good performance.
You know what you can buy with $900?
A lot of other essential things such as hookers and blow.
That's because measuring performance with FPS metrics is inaccurate. Have you seen this new frame latency data? If not, checkout TechReport.You can always match a high end setup for a cheaper SLI setup. A single card remains a far better solution than dual GPUs, as dual GPUs introduce a lot of problems, and microstutter can make a game feel like it has a low framerate even with good performance. My single GTX 680 feels faster than my old SLI GTX 570 setup, even though it's only maybe 50% faster than a single GTX 570.
Depends on if you do SLI or not. A 16x PCI-E slot should be enough for it I reckon. The SLI issue would come from the fact that a lot of motherboards have much less lanes for the additional GPUs.Can anyone confirm if I need to get a motherboard with PCI Express 3.0 to take advantage of that 85% performance increase? Or will my 2.0 do fine?
Depends on if you do SLI or not. A 16x PCI-E slot should be enough for it I reckon. The SLI issue would come from the fact that a lot of motherboards have much less lanes for the additional GPUs.
$899
$899
$899
$899
$899
You know what you can buy with $900?
A lot of other essential things such as groceries, college tuition fees, apartment fee, car monthly and insurance.
1) It might bottleneck on PCI-E 2.0. Just ever so slightly.Awesome, thanks. Do you think my Phenom II 965 would bottleneck?
Well it is like 10 WiiU's duct taped together.
Awesome, thanks. Do you think my Phenom II 965 would bottleneck?
I agree with this.Phenom II will bottleneck even gtx 660 in cpu intensive titles - if you are debating $900 gpu then i5 K series with mobo for <400 should be your first priority
Consistent frame times aren't going to be of any use if they're consistently low.I agree with this.
But so so folks are clear, it's not like there is a frame time limit based on a CPU that won't change if you get a better video card. It's a lot more dynamic than that.
The following 3DMark 11 which is supposedly ran with the GeForce Titan GPU scores X7107 marks in Extreme mode, you can compare it to the GeForce GTX 690 which has a score of around X6000 marks in the same benchmark.
The score is almost twice that of a single GTX 680
Don't know if posted yet but sound really impressive if true:
http://wccftech.com/nvidia-geforce-titan-performance-exposed-crushes-gpu-3dmark-11/#comments
Now confirm SweClockers sources that Geforce GTX Titanium is not only based on the same circuit board computing Tesla K20X but also contains as much memory. The upcoming monster model equipped with 6 GB of GDDR5 memory, which suggests that the video card is likely to have a memory bus of 384-bits.
There's one risk in buying it.
All 3 consoles are using ati architucture - that means games might favout 89xx series cards in benchmark so next gen ato could be getting close to titan in titles from 2014 onwards.
There's one risk in buying it.
All 3 consoles are using ati architucture - that means games might favout 89xx series cards in benchmark so next gen ato could be getting close to titan in titles from 2014 onwards.
Next Gen gpus are at half (less than half in terms of Durango) the power of a non-titan 680. A titan 680 is in a completely different league.There's one risk in buying it.
All 3 consoles are using ati architucture - that means games might favout 89xx series cards in benchmark so next gen ato could be getting close to titan in titles from 2014 onwards.
Next Gen gpus are at half (less than half in terms of Durango) the power of a non-titan 680. A titan 680 is in a completely different league.
Next Gen gpus are at half (less than half in terms of Durango) the power of a non-titan 680. A titan 680 is in a completely different league.
I didn't meant it like Titan will be too slow. What i meant is that 8970 might be nearly as fast for 500$ when it's released in H1 2013.
£700/800.jeeeeez. I want this! how much in the uk?
So is the Titan the GTX 780 essentially? Or is it the Titan, within its own naming convention and the GTX 780 is coming out down the line?