Are we getting close to a new Maxwell card now?

[Cloudfire777]

I found this interesting, but I don`t know if Im reading too much in too it.

For the last 3 months we have had records of shipments of parts of a GM204 graphic card. Either cooling parts, circuits, test boards etc.

But June 6th there was a GM204 graphic card shipping. A complete card I assume, and the value of the card is $550. Much higher than the parts that have shipped earlier.
I have also checked with previous entries of older cards, GK104 and GK107, and the quantities that are shipped =/= cards that will be sold. The numbers are way too low for that.

What does the "128M" stand for?

 

[Mand]

I was expecting something at Computex, but the deafening silence makes me think that they're close, but not close enough for them to kill off any remaining 700-series sales by announcing too early.

 

[NTMBK]

Doubtful we'll see anything before Q4.

 

[Cloudfire777]

According to several sites, upcoming Tonga is based on GCN 2.0 and will rival Maxwell in efficiency. Since we already know Tonga is atleast 2048 shaders (+more performace/shader vs GCN1.0), it will outperform GTX 750 Ti by a heft amount.

I wonder if Nvidia will respond with GM204 to that card. We think it will be 20nm, but what if its 28nm and a upgraded more power efficient version that will replace GTX 680 (GK104)?
Take GK107 vs GM107. Same power consumption but twice as fast despite both being 28nm.

And if we look at the die size increase for sticking to the same node,
GK107 118mm2
GM107 148mm2
+26%

GK104 290mm2
GM204 365mm2
Compare that with GK110 which is 550mm2 it will be quite a small chip even if its still on 28nm

 

[Ed1]

being that we have seen GM204 parts being shipped around, I bet it will be 28nm and they will be mid to high end but not very high as there not 210 parts (replacements for 780/780ti).

 

[tviceman]

I was expecting something at Computex, but the deafening silence makes me think that they're close, but not close enough for them to kill off any remaining 700-series sales by announcing too early.

Nvidia has never used computex to announce new GPUs.

According to several sites, upcoming Tonga is based on GCN 2.0 and will rival Maxwell in efficiency.

I know it's just speculation that has been floating around, and it definitely could all end up true, but I'd be absolutely floored if GCN 2.0 beat out Maxwell on perf/watt (assumming both are on the same node). It's certainly possible, and it if it does happen then it is because AMD integrated HBM onto the GPU die (rumored). I certainly don't know how feasible that will be from a cost vs. performance cost on 28nm (because of the complexity). Nvidia has been focusing extremely hard on mobile (hence perf/watt being a top priority), it would be kind of embarrassing if AMD beat them at their own initiative.

Everything is just kind of one big question mark right now. To me, Maxwell was a really surprising improvement over Kepler, given the manufacturing circumstances going on right now.

 

[3DVagabond]

Rival, in marketing speak, usually means not as good. In the same ballpark, not to bad in comparison, but it never means beats it.

 

[Cloudfire777]

What I meant was that GCN 2.0 is equal to Maxwell in efficiency yes.
If Nvidia can do it, then AMD too I`m sure.

Since Nvidia is already out with Maxwell, AMD is the one who needs to make a move. So Im guessing R9 M295X plus whatever desktop card (R9 275X?), will be AMDs answer to Maxwell. But it will be much faster than 750 Ti, which means Nvidia will have to respond again.

Thus why all the GM204 shipments have happened lately and "GM206" is nowhere to be found on Zauba but there are a lot of GM204 entries

GM107 > Tonga > GM204 > Pirate Islands > GM200 > WhateverAMDcallit
(GM206 and AMDs midrange is there somewhere in the middle)

 

[tviceman]

What I meant was that GCN 2.0 is equal to Maxwell in efficiency yes.
If Nvidia can do it, then AMD too I`m sure.

Certainly if one person (company) can do it, then so can another. The difference though is that Nvidia has hired engineers in the past two years while AMD has laid off engineers AND Nvidia has outspent AMD on GPU R&D. Also, Nvidia made no qualms quite awhile back about their focus being centered on efficiency. AMD had been ahead of Nvidia in that regard, but Nvidia's efforts are really, really showing through right now. That's why I said I'd be really floored.

 

[Mand]

What I meant was that GCN 2.0 is equal to Maxwell in efficiency yes.
If Nvidia can do it, then AMD too I`m sure.

I don't think it's a foregone conclusion that one company will be able to exactly duplicate the other in every effort.

 

[Cloudfire777]

I don't think it's a foregone conclusion that one company will be able to exactly duplicate the other in every effort.

They been doing it now for how long, 6+ years?
The "GPU race" is boring as hell. The GPUs from the two companies are so alike in performance its not even funny. No card is able to outperform a card from the other company by any amount that I can remember.

But we will see, maybe you are right. I`m not holding my breath though

Certainly if one person (company) can do it, then so can another. The difference though is that Nvidia has hired engineers in the past two years while AMD has laid off engineers AND Nvidia has outspent AMD on GPU R&D. Also, Nvidia made no qualms quite awhile back about their focus being centered on efficiency. AMD had been ahead of Nvidia in that regard, but Nvidia's efforts are really, really showing through right now. That's why I said I'd be really floored.

So you think Maxwell comes as a result of having more money for R&D? I do not disagree with you here, but from the quote above, the two companies have been so close over the years I`m starting to wonder where AMD gets the money to stay so competitive against a vastly richer company?!
No offence, but I would absolutely love to see Nvidia crushing AMD this generation with Maxwell. Just because we finally get a distance between those two. But I have doubts.

 

[moonbogg]

2014 is boring so far. Maybe we'll see some new mid range products late December.

 

[SinOfLiberty]

2014 is boring so far. Maybe we'll see some new mid range products late December.

A bit earlier

 

[tviceman]

So you think Maxwell comes as a result of having more money for R&D? I do not disagree with you here, but from the quote above, the two companies have been so close over the years I`m starting to wonder where AMD gets the money to stay so competitive against a vastly richer company?!

AMD's small die strategy several years back got them back in the game (with good perf/mm^2 and fast release cycles) while Nvidia started largely focusing much of their efforts on compute (GT200). Nvidia being late with Fermi also allowed AMD to catch up some more. However, since Nvidia's focus on efficiency (because of mobile), they have moved back into a secure leadership position and have accelerated their architecture release cycle when compared with AMD.

No offence, but I would absolutely love to see Nvidia crushing AMD this generation with Maxwell. Just because we finally get a distance between those two. But I have doubts.

I DO NOT want to see that. If either company has a significantly large advantage over the other, then we'll see sustaining higher prices.

 

[parvadomus]

I don't think it's a foregone conclusion that one company will be able to exactly duplicate the other in every effort.

GPUs are too easy, sadly. AMD is always trying to get Nvidia performance with smaller dies, they play with Nvidia literally. :awe:

 

[f1sherman]

GPUs are too easy, sadly. AMD is always trying to get Nvidia performance with smaller dies, they play with Nvidia literally.

:biggrin: :biggrin:

 

[RussianSensation]

I DO NOT want to see that. If either company has a significantly large advantage over the other, then we'll see sustaining higher prices.

Not only that:

1) GM204 would be milked even longer; and NV would purposely delay GM210

2) If NV has little to no competition for 2 years with Maxwell, they will be in no rush to launch Pascal on time when they could be milking "old 2-year old" Maxwell tech for $650+;

3) Severe decline in dGPU market share, lower ASP and profits for AMD are far more detrimental to its going concern. If AMD's APUs/CPUs/server CPUs were making them a lot of $, then sure they could afford hundreds of millions of losses in the dGPU space - but they can't.

--> More ideal imo is for AMD to win next generation by 10-15% in performance. This way AMD users will get the price/performance option with AMD, while NV users will get their eco-system without even higher prices. IAMD needs to really focus on the mobile dGPU space. That's where they lost the most mindshare and market share. If AMD gains market share in the mobile dGPU space, I would say, they could even afford to lose the single desktop GPU performance crown by 20% if their top card is $450-499 like R9 290X right now but NV can continue selling GM210 for $650-700. That would be fine for many PC gamers too. But, since Maxwell has shown such outstanding performance/watt improvement over Kepler on 28nm, I feel like next generation AMD will get owned in the mobile dGPU space.

 

[Flapdrol1337]

If nvidia has the better gpu's they'll simply charge more for them. Same with amd, maybe to a lesser extent, but neither wants a war of attrition.

 

[tviceman]

Not only that:

1) GM204 would be milked even longer; and NV would purposely delay GM210

2) If NV has little to no competition for 2 years with Maxwell, they will be in no rush to launch Pascal on time when they could be milking "old 2-year old" Maxwell tech for $650+;

Delay on GM210 for geforce, perhaps. We already saw that with Kepler and GK110, despite AMD being able to nudge out the gtx680 with the 7970GHZ and hold that lead until Titan (which was a really bad joke that, unfortunately, many people bought into). But I don't think diminished competition would slow down nvidia's roadmap. For the first time in Nvidia's history, all their products are (finally) sharing much of the same R&D (their gaming, mobile, and HPC are using the same GPU architecture). They're still trying heavily to compete in mobile and they need to stay aggressive to fend off Intel in HPC. It would just inflate prices enough along the geforce lines to slow volume without hurting the bottom line.

 

[CrazyElf]

Not only that:

1) GM204 would be milked even longer; and NV would purposely delay GM210

2) If NV has little to no competition for 2 years with Maxwell, they will be in no rush to launch Pascal on time when they could be milking "old 2-year old" Maxwell tech for $650+;

3) Severe decline in dGPU market share, lower ASP and profits for AMD are far more detrimental to its going concern. If AMD's APUs/CPUs/server CPUs were making them a lot of $, then sure they could afford hundreds of millions of losses in the dGPU space - but they can't.

--> More ideal imo is for AMD to win next generation by 10-15% in performance. This way AMD users will get the price/performance option with AMD, while NV users will get their eco-system without even higher prices. IAMD needs to really focus on the mobile dGPU space. That's where they lost the most mindshare and market share. If AMD gains market share in the mobile dGPU space, I would say, they could even afford to lose the single desktop GPU performance crown by 20% if their top card is $450-499 like R9 290X right now but NV can continue selling GM210 for $650-700. That would be fine for many PC gamers too. But, since Maxwell has shown such outstanding performance/watt improvement over Kepler on 28nm, I feel like next generation AMD will get owned in the mobile dGPU space.

Agree with this. Either a slight advantage for AMD or near parity. @OP It'd be horrible for us as end users for Nvidia to gain the advantage you speak of. Unless you enjoy paying more for fake prestige. Remember the Titan, the 8800 GTX, and when Nvidia has the advantage, they over-inflate prices. Heck, with a 15%-20% advantage, Nvidia would probably mark-up a lot. Remember, Nvidia rarely, if ever attempts to compete on price-performance. Quite the opposite - they tend to charge what they can get away with. With AMD with a 15-20% advantage, well, AMD has historically been much more aggressive with their pricing, so they may charge a premium but it'd be more in line with price to performance than Nvidia. That and AMD as noted above is in a difficult financial situation. We as a enthusiasts need two strong competitors. Not one. Otherwise we get horrible monopoly pricing.

 

[poohbear]

They been doing it now for how long, 6+ years?
The "GPU race" is boring as hell. The GPUs from the two companies are so alike in performance its not even funny. No card is able to outperform a card from the other company by any amount that I can remember.

But we will see, maybe you are right. I`m not holding my breath though



So you think Maxwell comes as a result of having more money for R&D? I do not disagree with you here, but from the quote above, the two companies have been so close over the years I`m starting to wonder where AMD gets the money to stay so competitive against a vastly richer company?!
No offence, but I would absolutely love to see Nvidia crushing AMD this generation with Maxwell. Just because we finally get a distance between those two. But I have doubts.

I dont see how u can say the 2 companies have been so close? Kepler is much more power efficient, runs cooler, & is quieter than the Radeons. How is that close? Performance wise maybe, but in terms of power efficiency, cooling, & noise nvidia's offerings are much better.

 

[wand3r3r]

I dont see how u can say the 2 companies have been so close? Kepler is much more power efficient, runs cooler, & is quieter than the Radeons. How is that close? Performance wise maybe, but in terms of power efficiency, cooling, & noise nvidia's offerings are much better.

Nope, you're actually mistaken. I assume you haven't been following it to closely.

Take a look at this.
http://www.techbuyersguru.com/sapphireR9290trix.php

 

[Unoid]

Nope, you're actually mistaken. I assume you haven't been following it to closely.

Take a look at this.
http://www.techbuyersguru.com/sapphireR9290trix.php

So AMD is at least equal Perf/watt to nvidia and at a cheaper price.

The poster above you just got told!

 

[tviceman]

So AMD is at least equal Perf/watt to nvidia and at a cheaper price.

The poster above you just got told!

Nothing comes close to Maxwell, and the vast majority of all published data shows GK110 being more efficient than Hawaii.

 

[wand3r3r]

Nothing comes close to Maxwell, and the vast majority of all published data shows GK110 being more efficient than Hawaii.

He specifically said kepler. I don't care that much for a few watts, however I just had to mention the data doesn't support his statements. Maxwell will be competing with GCN 2.0 too.