Follow along with the video below to see how to install our site as a web app on your home screen.
Note: This feature may not be available in some browsers.
It sucks that the wait has been so long, but really I think the problem here is manufacturing processes. AMD is fabless now, and while we could argue over whether or not that was a good decision, the delays on 20nm really impacted both NVIDIA and AMD's road maps.
Well fine then, there is no king of the hill coming anymore. It will just be a slower card than the titan x if u say so. Ill just go get 2x Titan X then
It sucks that the wait has been so long, but really I think the problem here is manufacturing processes. AMD is fabless now, and while we could argue over whether or not that was a good decision, the delays on 20nm really impacted both NVIDIA and AMD's road maps.
Nvidia knows they'll need a $700-$800 part to compete with the 390X which is already 10%+ faster than the Titan X. They can save a bit of cash on the VRAM. Factory OC closes the gap in benchmarks -- or maybe passes the 390X, which is what they're hoping for.geez, more rumors...
lol
an overclocked titanX with half the memory and the same TDP?
not a chance...
Nvidia's counter to the 390/X is in the pipe.
http://videocardz.com/55299/nvidia-geforce-gtx-980-ti-to-be-released-after-summer
Meh. The 780 Ti did this to the original Titan with 9 months inbetween. By the time the 980 Ti launches it will be virtually the same scenario.That would entirely invalidate the Titan X AND the 980, you'd have to be a moron to buy the Titan X when you get better performance from a card $399 cheaper.
That would entirely invalidate the Titan X AND the 980, you'd have to be a moron to buy the Titan X when you get better performance from a card $399 cheaper.
Nvidia has no choice but to be generous when they're competing with 8 GB Fiji's for the same price.If the 980 Ti really does come with 6GB, then the only reason to buy the Titan X would be for surround 4K gaming and that's it. Just seems way too generous on nVidia's part.
Nvidia's counter to the 390/X is in the pipe.
http://videocardz.com/55299/nvidia-geforce-gtx-980-ti-to-be-released-after-summer
That still leaves at least 5-8 months of waiting time in between before Pascal comes out, and what will they do with all the imperfect GM200 parts? Literally bin them?
I doubt it, i think the 980TI will have less shaders than the Titan-X as well as 6GB, I don't see them cannibalizing the Titan-X this soon,.
Has that ever happened though? I can't remember the last time nVidia rebranded their big dies as mid-tier cards in the succeeding lineup.
And these big dies are just far too hot and power hungry to convert to mobile parts, and nVidia hasn't tried that again since the 480M debacle.
The gutted the last Titan and Titan-Z with the 780ti. The one constant in technology is that it is ever evolving. I can't wait to see what Pascal can do. Didn't Nvidia's CEO say 10x the performance of the the 900 / Titan X series of cards in 2016? That makes Titan X stillborn on day one in my book. I mean it's really freaking nice for today, but spending $1,000 for 1 year of service is a bit much for my tastes.
680 became a 770 I think and the 670 became a 760. I mean they beefed up some of the tech attached to them, and made sure that SLi didn't work with the older cards. But they were effectively the same cards. I might be slightly off on the model numbers though. Sure that someone can correct me.
10x in what though? compute?
Anyway, they Did it with the titan / 780TI, which is why I don't see it happening again, the 780ti came out of nowhere, maybe as a knee-jerk response to AMDs 290/290x.
Anyhoo I guess we will see shortly, Also the 780ti had no DP/FP64 performance to speak of, so the Titan still had ( and still has) a useful place in the market, where as the Titan X would like you said, be stillborn.
Prime_1 linked this in the Nvidia forum.
http://hardforum.com/showthread.php?t=1855749
Bullets near the top mention it, but honestly since I'm not in the market for a Titan X, I paid little attention to the presentation.
http://blogs.nvidia.com/blog/2015/03/16/live-gtc/
The y-axis shows the potential throughput in arithmetic operations/second. The x-axis is operational intensity which is the number of operations applied to each data value (in units of operations per byte). The intensity is much lower for sparse operations e.g. sparse matrix multiply typically involves only a multiply and add for each input datum, while dense matrix multiply uses each datum many times. The horizontal lines reflect the maximum ALU throughput for each type of processor (the graph is drawn for Intel i7 and NVIDIA GeForce GTX 680 processors). GPUs have much higher ALU throughput since the GPU chip area is almost entirely ALU. For dense matrix multiply, GPUs are10x faster thanks to this higher computing area.
The diagonal lines reflect memory bandwidth. Since bandwidth is flow in bytes/second it defines a linear relationship between the x-axis (flops/byte) and the y-axis (in flops/sec). A less well-known feature of GPUs is their higher main-memory bandwidth. This leads to a (potential) 10x gap in sparse matrix operations, which are the most important for many machine learning tasks. CuSparse achieves this ceiling for typical scientific data, but we found it was less well fit to very sparse data (text, web logs etc.). We wrote our own sparse kernels and were able to get them close to the roofline limits over a full range of sparseness. These kernels form the basis for the high throughput in most of BIDMachs algorithms.
No you're correct in what you said, but notice how I said "big dies". GK104 was the mid-sized brother of the GK110 big die.
What I'm saying is if nVidia holds on to these GM200 big dies with defects and rebrands them in the upcoming series, that would definitely be a first for them.
Rebrand them as the low end solution for DX12 when Pascal comes out.
Has that ever happened though? I can't remember the last time nVidia rebranded their big dies as mid-tier cards in the succeeding lineup.
Yes, but both were sold (and priced) as high end parts. It's also debatable whether the vanilla Titan belongs in the 600 or 700 series. I think it's equally arguable the vanilla Titan marked the beginning of the 700 series, and served as its first release. In that sense, the Titan X could either be seen as an extension of the 900 series, or the beginning of the 1000 series, depending on what the next GM200 card will be called.
Regardless, if you follow the paper trail I was responding to this comment:
To which I said:
That was the context of me saying it would be a first for nVidia.
Again, 680-770 were both based on GK104, a mid-sized die.
480 was GF100, 580 was GF110, a more refined version of GF100, so it wasn't a straight rebrand.
And goddamn we've gone really OT.
So when the 390X comes out, the 980ti will be released not 2 months after which is full maxwell with 6GB and faster clocks than a Titan X and ~$300 less expensive. What is AMD gonna do then, assuming the 390X is even faster than a Titan X it's going to most likely get beaten by the 980ti with a lower price tag.
If all that is true, could that spell the doom for Red Team?
You're looking at this backwards.So when the 390X comes out, the 980ti will be released not 2 months after which is full maxwell with 6GB and faster clocks than a Titan X and ~$300 less expensive. What is AMD gonna do then, assuming the 390X is even faster than a Titan X it's going to most likely get beaten by the 980ti with a lower price tag.
If all that is true, could that spell the doom for Red Team?
I didn't mean problems with 28nm maturity. What I meant was that GPU designs are roadmapped and built for years in advance, and both NVIDIA and AMD planned to be on 20nm already by now. So basically every GPU design we've seen from AMD and NVIDIA in the past couple of years has been a compromise due to 20nm not being available, and you can be sure that some features or improvements have been changed or dropped to accommodate this. My guess is that they had to rework something to produce Fiji on 28nm. Plus if the rumors are true, they switching to GlobalFoundries for producing Fiji, and it can be a long process to get another fab up and running on your GPU designs.But the 390 is still on 28 just like the titan and the 28 process is about as mature as it gets. Seems blaming it on manufacturing issues at this point is far less likely than "oops overestimated 290 demand due to assuming bitcoin bubble would last forever and over produced them to the point that our distribution centers let employees take whole pallets home so their kids can use them to build GPU forts because we don't have enough storage to keep them at the warehouse"