So NVIDIA has announced their next generation of graphics processors, based on the Pascal architecture. They introduced it as “a new king,” because they claim that it is faster than the Titan X, even at a lower power. It will be available “around the world” on May 27th for $599 USD (MSRP). The GTX 1070 was also announced, with slightly reduced specifications, and it will be available on June 10th for $379 USD (MSRP).
Pascal is created on the 16nm process at TSMC, which gives them a lot of headroom. They have fewer shaders than the Titan X, but with a significantly higher clock rate. It also uses GDDR5X, which is an incremental improvement over GDDR5. We knew it wasn't going to use HBM2.0, like Big Pascal does, but it's interesting that they did not stick with old, reliable GDDR5.
The full specifications of the GTX 1080 are as follows:
- 2560 CUDA Cores
- 1607 MHz Base Clock (8.2 TFLOPs)
- 1733 MHz Boost Clock (8.9 TFLOPs)
- 8GB GDDR5X Memory at 320 GB/s (256-bit)
- 180W Listed Power (Update: uses 1x 8-pin power)
We do not currently have the specifications of the GTX 1070, apart from it being 6.5 TFLOPs.
It also looks like it has five display outputs: 3x DisplayPort 1.2, which are “ready” for 1.3 and 1.4, 1x HDMI 2.0b, and 1x DL-DVI. They do not explicitly state that all three DisplayPorts will run on the same standard, even though that seems likely. They also do not state whether all five outputs can be used simultaneously, but I hope that they can be.
They also have a new SLI bridge, called SLI HB Bridge, that is supposed to have double the bandwidth of Maxwell. I'm not sure what that will mean for multi-gpu systems, but it will probably be something we'll find out about soon.
I’ll get the GTX 1070 and
I’ll get the GTX 1070 and game on my 1080p 21.5 monitor and show a video doing it so I can piss people off how not to play games on a badass video card :-p
Holy shit this is big@!
Holy shit this is big@!
Thankyou.
Thankyou.
Glad to finally see 8GB VRAM
Glad to finally see 8GB VRAM on mid range card form Nvidia.
AMD Polaris for its own sake better go to 11 😉
Oh it goes to 11.
Oh it goes to 11.
AMD didn’t have 8gb on
AMD didn’t have 8gb on midrange cards either, 390 cards are not mid range even though people want to claim they are.
GTX600 = GTX1000 = scam
GTX600 = GTX1000 = scam
I don’t understand what this
I don’t understand what this comment is trying to say.
I dont know about the gtx600
I dont know about the gtx600 scam, but this in theory should cut the prices by a lot on 900series.
i just check the 4k benchmarks of Digital Foudry for 908ti vs titanx vs 980sli, and 1080 looks to me like a more power efficient 980ti. No more.
in the benchmark 980ti the OC version is equal or faster than titanx , 980sli won in most games, so if we put 980sli and 980ti OC on that slide it will show little performance gain.
So does the NDA lift before
So does the NDA lift before May 27th?
NDA lifts on May
NDA lifts on May 17th
https://forum.beyond3d.com/posts/1912730
Something doesn’t add up here
Something doesn’t add up here for the 1080. Micron still state on their website that DDR5X still isn’t in mass production so how can NVidia have enough chips? Sounds like very limited availability for the 1080 initially.
I guess they could have some kind of exclusivity deal? Who knows.. I guess we’ll find out soon enough.
Both the 1070 and 1080 look like great cards though. Kudos to Nvidia.
Paper launch isn’t new for
Paper launch isn’t new for this kind of luxury products…
Seems this may be a major
Seems this may be a major reason we have the expensive ‘Founders’ edition and no AIB until the middle of June at the earliest.
The founders edition 1080 at $699 means Nvidia have a monopoly for a month and will most likely make a profit selling reference cards for the first time since the GTX 680, amazingly Nvidia often lose money on reference cards.
This monopoly has given Nvidia extra time to squire a supply of GDDR5X to meet AIB’s needs, while the price itself slows sales allowing Nvidia to ensure they themselves have adequate GDDR5X supply for the 1080 Founders. This of course is not the only factor affecting price, the Founders MRSP is also clearly set high at $699 to politically appease AIB partners, allowing Nvidia a monopoly while minimizing any relationship damage or financial harm to AIB’s. Imagine how AIB partners would react if they lost millions in revenue directly due to Nvidia deciding to maximise month 1 sales and use the monopoly to offer the GTX 1080 at the normal lower reference card prices?
Regardless I have a funny feeling the Founders price of both the 1080 and 1070 is going to cause problems and possibly backfire – rightly or wrongly Nvidia understandably feel entitled to create the Founders cards due to the $2 Billion development cost of the new architecture, new 16nm process, new cards and new software – yet as the Founders cards are still nothing more than reference cards, how does Nvidia expect AIB partners to fill the price gap between the Founders MSRP $699 and lower MSRP $599? The only way to fill this price point is if AIB’s use cheaper components, substandard to the reference card – yet we all know most custom cards improve on reference specs, offer higher clocks, more power phases, better cooling etc. Nvidia really expects AIB’s to create cheap cards using cheaper than reference parts?
So either AIB’s such as ASUS, EVGA, MSI, Galax, invest R&D, and testing into creating cheaper cards substandard to reference or consumers dont see cards below $699. So either AIB image will be damaged by cheap product or consumer wallets will be damaged by Nvidia decision to create the Founders Edition.
I think the most likely outcome is sub $699 cards will not be seen for a long while, they will not appear until manufacturing processes are further refined – such as increased chip yields, and PCH development such as miniaturization, allowing greater efficiency with lower material costs without compromise in quality.
1733 MHz Boost Clock. Who
1733 MHz Boost Clock. Who else thinks we will be seeing 2Ghz OC cards in the next generation?
you should have looked into
you should have looked into this a bit more. They demonstrated a 1080 ostensibly doing 2.1 ghz. That’s with a 180 TDP from a single 8-pin power connector. That a max of 225 watts, 150 from the 8 -pin and 75 from the Mobo.
Not a ton of overhead for that kind of overclock. Some of the custom OC cards will have significantly more overhead for OC’ing. I think we could see N2 OC get near 3 ghz, which would be nuts.
You can’t base power
You can’t base power consumption on the connector layout; cards can violate the standard, and frequently do when overclocking.
All you do is point out the
All you do is point out the 295×2, that card would draw as much as 250watts from each 8pin it had. Card could draw up to 600watts with only 2x8pin.
i hope ryan focuses on 1080
i hope ryan focuses on 1080 vs furyx/390x on his benchs, instead of 980ti/X, because the state of maxwell drivers lately make me think it might have been gimped specifically for this purpose.
anyone else feel the same ? or am i reading too much into it ?
I can’t to see some
I can’t to see some benchmarks from all the youtube tech channels. I am most impressed by the fact that these cards have one 8 pin power connector and the performance that they put out.
Other than the price and the
Other than the price and the exact date of release there wasn’t much useful information given out, just another night of non-stop fluff. We already knew the names and new we’d see them in June. Wake me up when we get to see the real performance.
new=knew :L
new=knew :L
New series looks OK to me. I
New series looks OK to me. I still wanna see more proper drivers. Not this , will it work or not versions.
I’m excited to see this
I’m excited to see this proclamation of graphics power and only using a single 8-pin power connector. If this holds true at launch it would be amazing. Just going to the 16nm node would save power, and fin-fet does as well so I was expecting some vast improvements, but this power/performance, if it holds true, could very well be a game changer.
I’m still going to wait for Polaris, and reviews by people I trust (your trusted reviewers may be different) before making my next GPU purchase. I hope AMD does well, but they don’t have an anchor on my soul, maybe a drag-line though. (I’ll probably not be waiting for the ‘big’ versions of the chips at the end of the year/2017, the better price/perf with easy driving of 4k would probably get my money.)
The typical price gouging
The typical price gouging from Nvdia. Yay! I hope the GTX 9xx series drops in price. If there is a fire sale to move off the remaining inventory, I might just scoop up a 970.
Any idea if there will be a
Any idea if there will be a pre-order available for these cards and when that might start?
Yes. I’m taking pre-orders
Yes. I’m taking pre-orders now, no joke.
as far as the review goes id
as far as the review goes id specifically like to know more about:
-exactly how GPU Boost 3.0 compares to 2.0
-how the extra bandwidth of the sli bridge affects especially for non AFR modes (Explicit Multi-Adapter)
-To what degree over clocking is affected by water cooling (stretch goal haha)
I’m holding out for the GTX
I’m holding out for the GTX 1440 since I’m planning on buying a QHD gaming monitor soon.
Your lucky,
You don’t have
Your lucky,
You don’t have to wait as long as people wanting to play at 4320.
what is ” three DisplayPort
what is ” three DisplayPort 1.2 ports (1.4 “ready” with 4K at 120Hz) ” ?
this DP 1.2 or 1.4 ?
what is 1.4 “ready” with 4K at 120Hz ?
Also when are DisplayPort 1.4
Also when are DisplayPort 1.4 4k @120Hz HDR monitors coming out so we can use the full potential of this card since it’s DisplayPort 1.4 “ready”.
what with HDR ?
1000nits or
what with HDR ?
1000nits or 2000nits ? or ?
I’m building a new gaming PC
I’m building a new gaming PC right now and have collected all parts except for the GPU. I’ve hit the timing right since I didn’t have to wait too long for the new cards to come out. I’m partial to NVIDIA 1080, but I’m excited to see what AMD has up their sleeves too. Can’t wait!!!
Nvidia has better thread
Nvidia has better thread scheduling granularity now on its GP100 GPU micro-architecture, so things can be done at the instruction boundary instead of after the block level. How Pascal’s Asynchronous compute improvement will compare to AMDs GCN/Ace units’ asynchronous compute and AMDs asynchronous compute generational improvements will have to wait for the benchmarking software and enough DX12/Vulkan titles are on the market. Those DX12/Vulkan enabled gaming engines/Games will have to be on the market for at least a year before their DX12/Vulkan gaming engines are fully up to speed on these new graphics APIs.
Those P100 micro-architecture improvements(1) needs to be brought down into Nvidia’s consumer(GP104) SKUs if they are to be of use for VR gaming. AMD’s Polaris is coming for the consumer market first, with Vega being the big core in 2017, so AMD’s latest GCN improvement to its already in use for a few GCN generations hardware based Asynchronous compute will be getting more tweaks with Pascal.
For the HPC/exascale markets, I see AMD getting even more CPU like functionality into their ACE units, which will force Nvidia to add more CPU like functionality to their HPC/exascale SKUs. Intel, Or any other CPU maker, will have to keep an eye on both AMD and Nvidia if they both keep adding more CPU like functionality on their HPC/Server/exascale GPU accelerator SKUs, and those SKUs technologies will find their way down into the consumer variants because the VR gaming makers will want as much CPU types of functionality on Nvidia’s and AMD’s consumer SKUs to cut down on that CPU to GPU latency problem for VR gaming!
U-Sam is throwing billions of government funding into its Exascale computing initiative, so that funding to Nvidia, AMD, Intel, others will result in some very well funded R&D that will work its way from the Exascale computing market for GPU accelerators funding/R&D down into the consumer GPU, and CPU, markets!
(1)
“Compute Preemption is another important new hardware and software feature added to GP100 that allows compute tasks to be preempted at instruction-level granularity, rather than thread block granularity as in prior Maxwell and Kepler GPU architectures. Compute Preemption prevents long-running applications from either monopolizing the system (preventing other applications from running) or timing out. Programmers no longer need to modify their long-running applications to play nicely with other GPU applications. With Compute Preemption in GP100, applications can run as long as needed to process large datasets or wait for various conditions to occur, while scheduled alongside other tasks. For example, both interactive graphics tasks and interactive debuggers can run in concert with long-running compute tasks.”(1)
(1)
https://images.nvidia.com/content/pdf/tesla/whitepaper/pascal-architecture-whitepaper.pdf
edit more tweaks with
edit more tweaks with Pascal.
to: more tweaks with Polaris.