NVIDIA GM204 info is leaking

Subject: General Tech, Graphics Cards | September 8, 2014 - 05:03 PM |
Tagged: leak, nvidia, GM204, GTX 980, GTX 980M, GTX 970, GTX 970M

Please keep in mind that this information has been assembled via research done by WCCF Tech and Videocardz off of 3DMark entries of unreleased GPUs; we won't get the official numbers until the middle of this month.  That said, rumours and guesswork about new hardware are a favourite past time of our readers so here is the information we've seen so far about the upcoming GM204 chip from NVIDIA.  On the desktop side is the GeForce GTX 980 and GeForce GTX 970 which should both have 4GB of GDDR5 on a 256-bit bus with GPU clock speeds ranging from 1127 to 1190 MHz.  The performance that was shown on 3DMark has the GTX 980 beating the 780 Ti and R9 290X and the GTX 970 performing similarly to the plain GTX 780 and falling behind the 290X.  SLI scaling looks rather attractive with a pair of GTX 980 coming within a hair of the performance of the R9 295X2.

View Full Size

On the mobile side things look bleak for AMD, the GTX 980M and GTX 970M surpass the current GTX 880M which in turn benchmarks far better than AMD's M290X chip.  Again the scaling in SLI systems will be impressive assuming that the leaks that you can see indepth here are accurate.  It won't be too much longer before we know one way or the other so you might want to keep your finger off of the Buy Button for a short while.

Source: WCCF Tech

Video News

September 8, 2014 | 05:43 PM - Posted by Anonymous (not verified)

Ummm... clock-for-clock, no difference??!?

e.g. 1150/1127 = +2.04% for the 780Ti clock
12702/12328 = +3.03% for its benchmark score (ie 780Ti faster than its clock increase)

September 8, 2014 | 06:10 PM - Posted by Ashley (not verified)

They don't have a smaller process, so getting more IPC is difficult. My guess is they are focusing on lower power usage and features like Shadowplay, GSync, etc. That's what the R9 285 was and I don't see NVidia doing something VERY different from that.

September 8, 2014 | 06:43 PM - Posted by arbiter

well these are rumored numbers and they are on 3d mark firestrike so you can take it as a grain of salt. As for 285 being more power efficient, its only 10 watts which you could probably contribute to bus being cut down for the savings. AMD cards always had better results in firestrike compared to nvidia so.

September 9, 2014 | 05:06 PM - Posted by Anonymous (not verified)

Almost every 6xx card onwards has gsync and shadowplay so... wth.

September 8, 2014 | 06:40 PM - Posted by JohnGR

Is is rumored to be a 170W card. That's the HUGE difference.

September 8, 2014 | 07:11 PM - Posted by PapaDragon

Posted by JohnGR
Is is rumored to be a 170W card. That's the HUGE difference."

Exactly,then comes the price rumors. When all the cards came out from Nvidia(Titan,780ti..ect) and the prices settled. The GTX 680 remained at around $400.

When the GTX 800 (now 900 series) started to circle around , it was rumored to cost $399 because it was to replace the 680. Now its at $499 because its to replace the 780. So..Im really hoping, or I was, for that $399 price tag. .................But seeing how greedy Nvidia got with the Titan Z...Im certain it will launch at $499, hope not!

---Also, people have to look at the 750ti, its an insane overclockable card that you can find and buy at 1,200 mhz on the core and almost 1,300 boost. So expect great overclocks with non reference cards while having the tdp in check. Have you seen what the EVGA 780ti KingPin can pull??

September 8, 2014 | 08:45 PM - Posted by Anonymous (not verified)

We get that, but taking this as fact for a moment. We don't wait 1 1/2 years for a high-end card to simply have power improvements. It's nice to boast about the overclockability of a card, but it shows how overclocking is now being a selling point probably to artificially boost scores on review sites.

Both sides could clock cards higher allowing overclocking to still be possible, but harder to achieve the results we are now accustomed to seeing. When you can install your favorite software and make a couple clicks to get solid overclocks without even needed massive testing that just shows you the game they're playing.

I expect more from a next-gen card regardless. I don't care what process it's made on, that's not how it works. It never amazes me how fast people forget and just accept mediocre because its something new with a bigger number.

September 8, 2014 | 09:03 PM - Posted by Anonymous (not verified)

The highest overclocked 750 Ti still falls short of even coming close to a reference 660 performance.

What good is overclocking 100-200mhz if you only get 3fps out of it. I'd rather spend the extra $30 more and get a much better card.

September 8, 2014 | 09:37 PM - Posted by PapaDragon

Your right, but you're missing the point, the Gtx 660 tdp is 140w, the 750 ti is 60W, its meant more for HTPCs and low budget builds without consuming much, but still performs great.

Do remember that The 750ti has a measly 640 cuda cores on a 128 bit MIW. Not much to go on. The 980 is a completely different beast.

September 9, 2014 | 03:09 AM - Posted by Anonymous (not verified)

HTPCs don't need discrete gpus. On-board iGP is enough.

Unless you stick a god aweful PSU, a decent 400w could power most of the systems with up to a mid-range GPU in them. Unless your using a power brick laptop PSU there really is no concern.


119w with EVGA GTX 750 Ti ACX FTW
145w with Sapphire Radeon R7 265 Dual-X
152w with Nvidia GTX 660
175w with AMD Radeon R9 270X

30 watt difference from OC 750 TI to 660 and you still well under 200 watts system load during gaming for the cards listed.

I don't think 20-30watts is a crippling factor when the performance benefits are much greater with the other cards at a $10-$30 difference.

September 9, 2014 | 05:10 PM - Posted by Anonymous (not verified)

These are not laptop GPUs. I am buying GPU for performance. Up to 250W for desktop GPU setup is expected and fine by me, the only variables are performance and price. The former is unchanged according to leaks and thus the latter will be higher than 780ti (because it's new and AMD has not even announced any competing products).

September 8, 2014 | 06:18 PM - Posted by PhoneyVirus

It's pretty bad when the normal PowerUser like myself can predicted when Nvidia or AMD will come out with new graphic cards, and no I'm not talking about their tick tock strategy they got going on, when all you have to do is checkout TSMC.

Last year I said to myself, your lucky if you see something May or June 2014, even after they announced that they were launching around Q1 time frame. Then came TSMC, Nope Q2 Nvidia and AMD had to pull a Tick Tock all because TSMC couldn't get their gear up fast enough.

So comes Q3 2014 and we still get told the same crap, Q1 2015 okay so to TSMC that mean's late Q3, sometime between the ending of 2015 and the beginning of 2016, perfect let me know when the real graphic cards come out please. Not you guys PCPer.


September 8, 2014 | 06:19 PM - Posted by Anonymous (not verified)

my 780s are last gpus for me pc gaming is joke with crappy ports

September 8, 2014 | 07:36 PM - Posted by arbiter

can't blame the gpu if the game is a crappy port job.

September 8, 2014 | 10:14 PM - Posted by Anonymous (not verified)

never did

September 14, 2014 | 08:23 PM - Posted by Anonymous (not verified)

Totes agree, new games( console ports) aren't pushing gpu's. I'll probably upgrade to the 980ti SLI and the be done with it. Same goes for processors and RAM, no need to upgrade anymore. The GPU makers should be putting pressure on LAX game devs.

September 8, 2014 | 06:21 PM - Posted by GutsNotDead (not verified)

Charlie Demirijian wrote back in 2012 claiming that Nvidia would be using a more a power efficient 28nm process than the one AMD used to built their chips. If that is correct, then it's clear Why Nvidia's been successful in building more power efficient chips than AMD since then. Looking at those clocks of those 900 series cards kinda tells you how all that works out.

Now what matters more is the price. I don't care about lower power consumption on load that much, because I rarely game these days. Now AMD's new cards consume less power when the display is off, and my desktop machine spends more time in that state than gaming, so I hope the Green Goblin is sensible enough to price their stuff properly this time around.

September 9, 2014 | 11:38 AM - Posted by Luthair

Pretty big stretch given they both use TSMC, and if TSMC had some super duper 28nm process available they'd certainly be blabbing about it even if they were doing some exclusivity deals.

September 8, 2014 | 06:47 PM - Posted by JohnGR

The performance of the mobile chip shows that Maxwell is a chip perfect for low wattage.

Now on the desktop we will have to wait the 384bit models and what the chip can do on a 250W power. 980 is rumored to be only 170W.

That 256bit bus makes me believe also that this is more like a 770 successor and not a 780 successor. Nvidia is using the X80 brand to move the prices higher. I am expecting not just a 980Ti but also a 985/990 single chip in the future.

So the performance sector is moving higher by $100 with this generation.
How nice if true.

September 8, 2014 | 07:18 PM - Posted by GutsNotDead (not verified)

Aren't you expecting too much?

September 8, 2014 | 07:45 PM - Posted by arbiter

if its 170watts could leave a lot of overclocking headroom like 750ti seems to have. 750ti started at 1000mhz, some overclocks can get a stable 1400+ outta it and that was without a 6pin PCI-e power on a lot of them.

Well when you consider 750ti can do almost same work as 2 AMD parts it competes against which is 260x/265. Yes 750ti is little slower then 265 but can keep pretty close pace using only 60 watts. 260x is rated at 115watts, and 265 is rated for 150watts. I was figuring they could get it around 200watts or so but 170 for it would nice.

September 9, 2014 | 03:02 AM - Posted by collie

I often wonder the same ish thing about haswell. they can do some crazy low powered shizzle at 35w, what could they do if they went to 220w tdp on haswell? Possibly the architecture would perform pour at the high power/heat level, like prescot back in the day, Higher power=less power. Or maybe, and I'm sure the only people who know the answer to this already work for intel, maybe if they could push the TDP, trusting the motherboard side, trusting the cooling side, perhaps it could be sickness
Or not

I have to stop posing drung off my ass

September 8, 2014 | 09:15 PM - Posted by Anonymous (not verified)


September 8, 2014 | 10:55 PM - Posted by Anonymous_not t...

Wait wait wait lol... Sounds like you're still just a little bit too interested.

September 9, 2014 | 05:21 AM - Posted by Alamo

boring pretty much 290X perf, 170 watt i doubt it, anything from tonga or maxwell is gonna be boring, we are all stuck waiting for 20nm, stfu Apple, stfu TSMC, all of this is because of them!

September 9, 2014 | 10:45 AM - Posted by Anonymous (not verified)

Leaking? Hmm, this is more like the web version of premature ejaculation. Everyone is so desperate for tech news that they are blowing their collective loads on made up benchmarks. Let's assume they end up being accurate... so what? We can't buy them until October anyway.

September 9, 2014 | 04:03 PM - Posted by Lithium (not verified)

Pity that they didn't allow SLI on 750 Ti.
Triple 750 Ti SLI...hehe

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Lines and paragraphs break automatically.
  • Allowed HTML tags: <a> <em> <strong> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote><p><br>
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

This question is for testing whether you are a human visitor and to prevent automated spam submissions.