Tuesday, September 16th 2014
Even More GeForce GTX 980 and GM204 Specs Tumble Out
Ahead of its launch later this week, even more details of NVIDIA's upcoming GeForce GTX 980, and the 28 nm "GM204" silicon it's based on, tumbled out. To begin with, the GM204 silicon is confirmed to be built on the 28 nm silicon fab process. The chip bigger than that of the GK104, with a die area of 398 mm², yet smaller than the GK110, which measures 581 mm². Its transistor count is 5.2 billion, about 2 billion more than the GK104.
The component hierarchy of GM204 is similar to that of the GM107 silicon, on which the GTX 750 Ti is based. The GPU features a 256-bit wide GDDR5 memory interface, and PCI-Express 3.0 x16 bus. The GigaThread Engine dispatches workload between four graphics processing clusters (GPCs), the basic subunit. Each GPC has a common raster engine shared between four streaming multiprocessors Maxwell (SMMs), which each hold 128 CUDA cores. The total CUDA core count is hence 2,048. The L2 cache has been quadrupled over GK104. The chip features 2 MB of it, compared to 512 KB on its predecessor. The GM204 features 64 ROPs, double that of the GK104, and should hence come with a strong geometry processing muscle. The chip features a revolutionary new 3-bit delta color compression technology that makes the most of the limited memory bus width of this chip.
Here are the final specifications of the GTX 980 and GTX 970, carved out of this chip.
GeForce GTX 980
Source:
VideoCardz
The component hierarchy of GM204 is similar to that of the GM107 silicon, on which the GTX 750 Ti is based. The GPU features a 256-bit wide GDDR5 memory interface, and PCI-Express 3.0 x16 bus. The GigaThread Engine dispatches workload between four graphics processing clusters (GPCs), the basic subunit. Each GPC has a common raster engine shared between four streaming multiprocessors Maxwell (SMMs), which each hold 128 CUDA cores. The total CUDA core count is hence 2,048. The L2 cache has been quadrupled over GK104. The chip features 2 MB of it, compared to 512 KB on its predecessor. The GM204 features 64 ROPs, double that of the GK104, and should hence come with a strong geometry processing muscle. The chip features a revolutionary new 3-bit delta color compression technology that makes the most of the limited memory bus width of this chip.
GeForce GTX 980
- 2,048 CUDA cores
- 128 TMUs, 64 ROPs
- 256-bit wide GDDR5 memory interface
- 4 GB standard memory amount
- Core clock speeds of 1126 MHz, with 1216 MHz GPU Boost, and 7012 MHz memory
- 165W TDP
- 1,664 CUDA cores
- 112 TMUs, 64 ROPs
- 256-bit wide GDDR5 memory interface
- 4 GB standard memory amount
- Core clock speeds of 1051 MHz, with 1178 MHz GPU Boost, and 7012 MHz memory
94 Comments on Even More GeForce GTX 980 and GM204 Specs Tumble Out
i have another logic... because Friday is the most waited day for many, perhaps their attitude would be much more welcome and they will accept the product more warmly thus this will ensure even better demand, etc. ;)
stackoverflow.com/questions/6163549/how-do-services-like-dropbox-implement-delta-encoding-if-their-files-are-stored
Yet. let me use logic. NVIDIA's first global gaming event, game24, starts Thursday, the 18th. I am no marketing expert, but I will bet a 3$ bill they planned the GTX900 series announcement for that event.
Bear in mind that the GM 204 replaces the GK 104 (1536 shaders) in the vendors hierarchy, not the GK 110.
G80 -> GT200 -> GF100/110 -> GK110 -> GM 200
G92 -> M.I.A. -> GK104/114 -> GK104 -> GM204
New GPU in the same general performance range doesn't automatically mean it replaces the SKU's. I don't think anyone is expecting the fully enabled Tonga (285X) Volcanic Islands to replace the Hawaii (Sea Islands) GPU are they?
What I'm interested is if they have actually been able to improve the voltage and thermal capabilities of the architecture, in that case, these low power density chips will really be able to take advantage of higher current VRM solutions.
It is almost like rebranding. Very very bad and misleading. One needs to monitor very closely not to be misled by these numbering shenanigans.
Until very recently, everything with numbers was ok, now it came as a dirty idea in their minds. Wondering why they waste their times and don't think of something better.
In an ideal, honest and good world, these GM204 would be put in GTX 860 Ti, 860, etc. But no, it is in 980 and 970.
2. The 700 series were a mix or architectures (Fermi and Kepler). The R5/R7/R9 series from AMD is also a mix of architectures - not just Volcanic Islands and Southern Islands but the 4.5 year old Evergreen series and the 3.5 year old Northern Islands - yet Nvidia are greedy and marketing frauds and you've never said a word against AMD's conglomeration of architectures or their "skipping" of the HD 8000 series. Quite an oversightYet you choose to overlook AMD's rebranding. Is rebranding suddenly gaining importance for you personally over the last few months?......this is shades of your diatribe regarding the importance of double precision which suddenly became disinterest when AMD cut the rate from 1:4 in Tahiti to 1:16 in Tonga.The consumer doesn't win in any event. What is the difference between a whole new naming nomenclature with the same silicon ( 9800GTX+ -> GTS 250, HD 7970 -> R9 280X) or an incremental change in numbers? There is no substitute for doing your homework before making a purchase, and there certainly isn't any truth to the rumour that marketing is designed to make the facts crystal clear to the consumer.Probably reflects the economics of the time. Discrete graphics card sales are down, and process costs will go up. It seems likely that Maxwell (and AMD's Tonga etc) will both be die shrunk on TSMC's 16nmFF process, so getting the highest performance for the yield without pricing themselves out of the market will be the order of the day.
Large dies tend to pay for themselves through professional board sales, but that is a finite market so it becomes very unproductive to apply a large GPU down the consumer product stack.
And they are very stupid for doing so.
First because actually not their employees but the customers are their greatest asset, and second, because they can up the quality of what they offer to the maximum so customers are not forced to buy the product and at the same time to curse the company for the price.
When giving higher quality, they will still improve over generations, and will look better in our eyes. The standard itself will be higher.Normal, how do you expect sales to climb after what I'm telling you. Customers are not happy and it will get worse.
nvidia will go blindly in their way because they say so. :rolleyes:
Do you think that customers will not buy with the same 'success' a $599.99 GTX 860 Ti compared to the same $599.99 GTX 980? :D
Theoretically, both companies could put out their most cutting edge products - but once you start, ANY slip in timetable brings criticism, and ANY part failing to measure up to any perceived metric previously reached will start getting the haters out in force. 4K is barely relevant as a statistic in usage, yet a significant portion of forum users across the net are whining and moaning about the lack of 5K and 8K support.
Give the consumer what they want and they just want more, and as for tech "enthusiasts", I've yet to meet any significant number that are content with ANY aspect of technology. You'd be hard pressed to find a bigger bunch of self entitled ramblers anywhere. Give people a 150W card that does the job of a 250W card from the previous generation and they'll whine that it should be sub-30W and passive, give people HDMI 2.0 and they'll whine about (possibly) no DisplayPort 1.3 connectivity... or the card not being single slot, not shiny enough, not the right colour, not quiet enough, not cool enough, too big, too small (Hey! that doesn't look like a premium card!), not enough accessories they'll never use anyway - the list is endless. Like it or not, multi-billion dollar companies don't cater to a miniscule percentage of the market, especially one that will never be satisfied. You could provide a card that is butter smooth on a demanding title at 4K with 8 x SSAA/DoF/motion blur/HDAO applied, and the first thing out of a lot of peoples mouths would be a complaint that the company had previously been holding back, or the driver package is too large, or the software utilities aren't meaningful enough, or most likely - that the company cheated you because you can't push 2 volts through it. :laugh:
It's been this way for a while. The number of people of confuse "tech enthusiast" with "hyper-critical user with unrealistic expectations" seems to grow larger every product cycle. Intel could no doubt transform their 18-core E5-2699v3 into a desktop part, but how would they continue to keep topping something like that? The moment they slip up and "only" produce a 30-core next time around you have a bunch of people complaining because the rate of progress isn't to their satisfaction - basically back to square one. Sometimes, you just have to apply real world values rather than a wish list - businesses are about sustainability. and they'd soon go out of business if they had to come up with a new process node every year and offer cards for relative peanuts when they have to fork out $7+K per wafer.The major reason discrete board sales are down is because of iGP's. Not everyone games, and of those who do, a fair percentage are casual or playing some inane flash PoS. Once upon a time integrated graphics were just a method for OEMs to cheap out on finding away to display video. That isn't really the case anymore when the vast majority of desktop processors include graphics - that is why sales are falling.
It's not like we have games or anything that really truly requires more, consoles are stagnant for a few years, and the only thing that can push GPU needs in the consumer market is higher-resolution monitors, and 3k/4k monitors have yet to gain traction. Looking at this from a business perspective, it's pretty exciting.