Nvidia's Pascal GP104 GPU may opt for GDDR5 over HBM *Updated GDDR5X JDEC Specs

[MarkizSchnitzel]

Since that's what the 980 (GM104) came out at. I don't expect 16nm FF GP104 to be any cheaper.

We have different ideas of what midrange is
200-300$ is midrange.
More than that is enthusiast/high.

 

[Qwertilot]

Prices move too much over time to be used to define that sort of thing

The absolutist way of defining mid range is simply in terms of power draw - there will eventually be next gen stuff at the same sort of TDP points as the current stuff, and the performance gaps between them will be about the same as it is nowadays.

We'll see in 6-9 months or so.

 

[zlejedi]

Give us the cards and we will judge design decisions with benchmark results in hand I don't care what tech they use as long as they bring results.

 

[3DVagabond]

We have different ideas of what midrange is
200-300$ is midrange.
More than that is enthusiast/high.

That was before the high end was $650-1000. Besides, we're arguing semantics.

 

[Goatsecks]

If you had a 980 ti would you pay $550 just to use less power with no visual improvement?

Sorry, what is $550? The 980ti? I am confused.

 

[tweakboy]

When is it coming out ...... Im waiting for the 980 Ti successor ..... sorry spelling,, thanks

 

[prtskg]

Sorry, what is $550? The 980ti? I am confused.

He is expecting that the price for a next gen card with performance slightly ahead of 980ti will be $550.

 

[2is]

If you had a 980 ti would you pay $550 just to use less power with no visual improvement?

Obviously if the performance wasn't any better, it would not be marketed towards current 980Ti owners. You say that like everyone owns a 980Ti.

 

[Goatsecks]

He is expecting that the price for a next gen card with performance slightly ahead of 980ti will be $550.

Obviously if the performance wasn't any better, it would not be marketed towards current 980Ti owners. You say that like everyone owns a 980Ti.

Wow Vagabond, i'm surprised you didnt give yourself a hernia straining that hard.

 

[tviceman]

If you had a 980 ti would you pay $550 just to use less power with no visual improvement?

You think next gen "flagship cards" are going to start at $550. Hahahaa.

 

[Silverforce11]

You think next gen "flagship cards" are going to start at $550. Hahahaa.

It's a bit sad we're expecting mid-range chip SKUs to cost way above $550...

 

[guskline]

From the interview I saw with the AMD rep the lower end Polaris cards will use GDDR5 ram vs HBM. I suspect the highest end cards will use HBM2.

 

[3DVagabond]

You think next gen "flagship cards" are going to start at $550. Hahahaa.

When they release the midrange as highend again, yes.

 

[nenforcer]

The question I have once again is why Jen-Hsun Huang was showing a Maxwell based GPU on the Drive PX 2 at CES which is suppose to feature a Pascal GPU. For a product that is suppose to be shipping in less than 6 months it doesn't look very promising. Very much a throwback to the Fermi GPU shown 5 years ago in 2009 that didn't release until 2010.

AMD has a real opportunity here with both an manufacturing switch down to .14nm and a memory technology change from GDDR5 to HBM2. They already have proven silicon for HBM1 and this new HBM2 flagship which doubles the memory bandwith again could allow AMD to retake the high end.

 

[2is]

When they release the midrange as highend again, yes.

Yep, that can happen if the same set of circumstances takes place. Those being, NVidia seeing that AMD's flagship card is slower then their mid-range cards. That's a caveat you are either unaware of or are intentionally leaving out. At that point who can blame them. Not to mention, not only was the 680 faster than the 7970, it was also cheaper. Sure, AMD bounced back with the GHz edition and driver updates that improved performance the combination of which made it faster than the 680, but that was a few months later.

 

[RussianSensation]

Yep, that can happen if the same set of circumstances takes place. Those being, NVidia seeing that AMD's flagship card is slower then their mid-range cards. That's a caveat you are either unaware of or are intentionally leaving out. At that point who can blame them. Not to mention, not only was the 680 faster than the 7970, it was also cheaper. Sure, AMD bounced back with the GHz edition and driver updates that improved performance the combination of which made it faster than the 680, but that was a few months later.

The above has nothing to do with the market dynamics he brought up. Almost every next gen mid-range card outperformed last generation's flagship. Go back to GeForce 2 if you want and check. His point was that the new way generations will be sold is next gen's mid-range will be marketed as a $550-600 flagship even if the chip's performance of that generation is mid-range, even if the code name is mid-range (GP104/204, etc.). How Polaris 300-350mm2 stacks up exactly against GP104 is more or less irrelevant to his point. He isn't discussing AMD vs. NV but the market accepting 2-3 "flagships" in the same generation.

Also, it doesn't matter at all how many people own a 980Ti/Fury X to make his point valid. For instance, the current 2017 C-Class or 2017 Audi A4 have way nicer interiors, safety and technology than the 2016 E-Class or 2016 Audi A6. Should they "temporarily" be marketed as 2017 "E-Class" and 2017 "A6" with $20K price hikes until the real next gen 2017 E-Class/2018 A6 show up?

With the GPU market since 2012, this is exactly what happened but a large majority of PC gamers do not care, do not know, or don't want to admit to this fact. With this new GPU gen era, AMD/NV only have to beat 980Ti by 15-20%, and they will sell it for $550-600. If I were NV/AMD I would just price it at $650 and say it's faster than 980Ti so it's a fair price.

Without getting into technical details of next gen 2016 cards, if they aren't 50-75% faster at 1440/4K than 980Ti and aren't using 220-250W of power, they aren't flagships no matter what marketing spin AMD/NV put. Go back to 1998 until now and you will see that with a node shrink+new architecture, performance went up 70-80%, easily. This gen we have HBM1/2 + FinFET + 14/16nm. That means if GP104 is only 20-25% faster than 980Ti, uses GDDR5/X, what is it? It's a mid-range Pascal chip in the Pascal generation.

 

[Goatsecks]

When they release the midrange as highend again, yes.

With a bit of luck Polaris will be good enough to oblige them to release a propper flagship this time round.

 

[3DVagabond]

With a bit of luck Polaris will be good enough to oblige them to release a propper flagship this time round.

I'm looking for the motivation for either company to do that. If you can release what should be a ~$300 card for $550, why not. I think AMD will do the same.

 

[Goatsecks]

I'm looking for the motivation for either company to do that. If you can release what should be a ~$300 card for $550, why not. I think AMD will do the same.

Are you implying collusion?

 

[3DVagabond]

Are you implying collusion?

Not at all.

 

[maddie]

I'm looking for the motivation for either company to do that. If you can release what should be a ~$300 card for $550, why not. I think AMD will do the same.

Assuming both companies pursue this strategy, with closely competing products. Where do you see AMD's market-share for 2016-2017?

 

[3DVagabond]

Assuming both companies pursue this strategy, with closely competing products. Where do you see AMD's market-share for 2016-2017?

That depends on execution by both companies. If they both release at the same time with similar product I see AMD doing better than they are but not getting equal market. If AMD outperforms nVidia they could get back to 35%-40% like they've had in the past. If nVidia is late (which might be the case seeing as how AMD has shown working silicon and nVidia hasn't) then they have a chance to maybe get even. If nVidia releases first or has better performance then AMD will have to give their stuff away just to sell cards.

 

[Head1985]

256-bit bus with 12ghz GDDR5X = 384 gb/s, a 70% increase in bandwidth over GM204. 256-bit bus would be just fine.

A 70% increase in performance (guessing) along with a 70% increase in bandwidth? Interestingly, that would make it only about 7-12% faster than the Gigabyte water-cooled GTX 980 TI in 1440p and 4k. Hopefully there is plenty of headroom in Pascal or people with overclocked 980 TI's will be laughing at how small the performance gain would be for the next "high end" sku from nvidia.

But gigabyte GTX980TI waterforce already runing at 1430Mhz!If GP 104 at stock is 10% faster than 1430Mhz GTX980TI that pretty good.

Stock vs stock GP 104 should be 40% above TITANX(50% above 980TI)

Btw 980TI waterforce is 51% faster than GTX980.If GP 104 is 70% faster than GM204 we will see increase 20% over 980TI water force and 40% above TITANX and 50% over stock 980TI.

 

[Headfoot]

If anyone here is seriously suggesting that nVidia and probably AMD won't release a middle sized chip as a flagship and charge a flagship price once 14+16nm launch, I've got a bridge to sell. Real cheap

 

[Techhog]

We have different ideas of what midrange is
200-300$ is midrange.
More than that is enthusiast/high.

He's talking in talking in terms of place in the chip lineup. The 980 is based on the midrange Maxwell chip. This is a fact, You can consider it a high-end card, but that won't change the fact that it's not big Maxwell. The same went for the 680. We saw full GK104 at $500, and we saw full GM204 at $500, so it's safe to say that full GP104 will be sold at $500 or more.