GeForce 20 series to be released this year (Volta)

razor1 said:
https://videocardz.com/68948/skhynix-gddr6-for-high-end-graphics-card-in-early-2018

SK Hynix has just confirmed that high-end graphics card based on GDDR6 technology will be available in early 2018.

NVIDIA Volta with GDDR6 in early 2018?
Click to expand...

Maybe for their upper GPUs for now (HBM may just be for their products such as GP100/GV100/etc) but comes down to what also AMD does and separately when it can be tested within product cycle.
Nvidia usually buys their memory these days tech from Samsung/Micron.
And lets not forget how enthusiastic SK Hynix were in over promoting their availability of HBM2; will take a bit of time for GDDR6 to hit its full potential in terms of spec.
Cheers
 
Last edited:
CSI_PC said:
Maybe for their highest GPUs but comes down to what also AMD does.
Nvidia usually buys their memory these days tech from Samsung/Micron.
And lets not forget how enthusiastic SK Hynix were in over promoting their availability of HBM2.
Cheers
Click to expand...

Hynix have to step up tho, else they are becoming irrelevant to say it mildly in the graphics area. I also bet you that's why this PR statement is there. Even for HBM2 they lost with a year.

But you are right, Micron, Samsung and Hynix will all deliver GDDR6, that again is really just a standardization of GDDR5X.
 
Shintai said:
Hynix have to step up tho, else they are becoming irrelevant to say it mildly in the graphics area. I also bet you that's why this PR statement is there. Even for HBM2 they lost with a year.

But you are right, Micron, Samsung and Hynix will all deliver GDDR6, that again is really just a standardization of GDDR5X.
Click to expand...

Yeah.
It will also be interesting to see what rating all of them can initially launch with, still feel Samsung or Micron will have a viable product available before SK Hynix and your right SK Hynix are under pressure in this segment.

Cheers
 
SomeGuy133 said:
Shintai makes shit up all the time. Pay little attention to his claims. I have called him out on his BS several times



those rumors never claimed consumer. They were broad blank statements on volta...so not consumer.

.
Click to expand...
Actually it was consumer.
V100 has been known for awhile and it is mid-late summer.
Volta Tegra has been known for awhile and that is Q3/Q4.
That leaves consumer for the rumours that those same publications also used to say 2018 for Volta (so it is not a broad blank statement), but even then there was a good reason for at least one consumer product in Q4.

Cheers
 
CSI_PC said:
Actually it was consumer.
V100 has been known for awhile and it is mid-late summer.
Volta Tegra has been known for awhile and that is Q3/Q4.
That leaves consumer for the rumours that those same publications also used to say 2018 for Volta (so it is not a broad blank statement), but even then there was a good reason for at least one consumer product in Q4.

Cheers
Click to expand...
you mean voltage in the tegra SOCs right? That has nothing to do with consumer desktop and everything to do with server so my statement is still accurate.
 
SomeGuy133 said:
you mean voltage in the tegra SOCs right? That has nothing to do with consumer desktop and everything to do with server so my statement is still accurate.
Click to expand...
Remember Tegra is highly modular.
Volta Tegra is also Xavier (Q3/Q4 2017 sampling with key manufacturers), last year Pascal Tegra was Drive PX2 (did not go into sampling until late 2016).

But you ignored my whole post I notice.
It is important because the rumour cannot be the Tegra nor Tesla HPC, ergo that leaves consumer with the latest Q3 rumour :)
And those same sites also said Volta 2018, well obviously they were not talking generally about Volta for the reasons I explained as we know the cycle for Volta HPC and Tegra already for 2017.

You said:
those rumors never claimed consumer. They were broad blank statements on volta...so not consumer.
Click to expand...
It cannot be a broad blank statement because it ignores the 2017 launch of the HPC-server-etc (that was not their focus for such news).
Cheers
 
Last edited:
CSI_PC said:
Remember Tegra is highly modular.
Volta Tegra is also Xavier (Q3/Q4 2017 sampling with key manufacturers), last year Pascal Tegra was Drive PX2 (did not go into sampling until late 2016).

But you ignored my whole post I notice.
It is important because the rumour cannot be the Tegra nor Tesla HPC, ergo that leaves consumer with the latest Q3 rumour :)
And those same sites also said Volta 2018, well obviously they were not talking generally about Volta for the reasons I explained as we know the cycle for Volta HPC and Tegra already for 2017.

You said:

It cannot be a broad blank statement because it ignores the 2017 launch of the HPC-server-etc (that was not their focus for such news).
Cheers
Click to expand...
still dont see 2080 or titan end of this year. I'll believe it when i see official news. I highly doubt it.
 
SomeGuy133 said:
still dont see 2080 or titan end of this year. I'll believe it when i see official news. I highly doubt it.
Click to expand...

No-one has said Titan as that launches a fair bit after the 1080/2080 and a likely candidate for GDDR6.
The only GPU likely in Q4 on the consumer side would be GV104.

You realise Nvidia has to keep quite tight synergy between their diverse segment for cost and logistical aspects of R&D/manufacturing, and ask yourself why Nvidia did one of the fastest and tightest product cycle rollouts in their history with Pascal across all technology segment if you prefer not to speculate on the other factors previously given by a few in the past for why Q4 is a high possibility for one-two consumer Volta GPUs.
Cheers
 
Last edited:
SomeGuy133 said:
still dont see 2080 or titan end of this year. I'll believe it when i see official news. I highly doubt it.
Click to expand...


nV has been hitting their new gen per 1.5 year time line since Fermi, no need to doubt that now. Titan for consumer most likely won't be EOY, consumer (gv104) EOY or early next year.
 
NKD said:
I am scratching my head here. When we had this conversation before you said late 2017 wil be 1.5 years and that is how nvidia goes. Now early 2018 is 1.5 years. How does that exactly make sense? It will be almost 2 years. There is nothing wrong with that and it makes sense since nvidia would have Volta out for deep learning field like they did with pascal and 6 months or so later they will announce high performance consumer Volta parts just like they did with pascal. So we are looking at 2 year upgrade cycle now. Late 2017 is 1.5 years not early 2018.
Click to expand...


January of 2018 still is close to 1.5 years, a month or two deviation can be understood.
 
It also needs to be considered Pascal was a technical risk/milestone towards Volta as it introduced certain technologies with a node shrink while the rest will come with Volta or be further evolved (such as NVLink 1 to NVLink 2).
The way Pascal launched was anything but usual, because it is a forerunner for Volta.

GV100 is expected 15-16 months after GP100.
Early Q4 will mean 15-16 months between launch of GP104 and GV104.
Volta Xavier (Tegra) is on a similar or accelerated schedule (meaning the Volta architecture is well worked out and operational), this usually launches after both the initial Tesla and some consumer products with the same architecture.

TBH I thought it would be late Q4 for consumer (only for GV104 this year), but the option is there if Nvidia wants to be aggressive with this product initial cycle rollout across all segments earlier that the very latest rumours hint at albeit surmising the reason was consumer GPU sales down (we know Tesla-'Server' is through the roof in growth relative to consumer), and we know they can and do this even without competition from AMD (ala Pascal).
That said more recent analysis suggests sales for consumer GPU are still stable, so if Nvidia did move this forward I would think it relates to other reasons.
Cheers
 
Last edited:
razor1 said:
January of 2018 still is close to 1.5 years, a month or two deviation can be understood.
Click to expand...

You really expect it to drop January? The month the memory goes to mass production? Earliest will be march-may. Looks we we will have the same time line pascal. May be little early. But same timeline as pascal seems to be the thing here.
 
NKD said:
You really expect it to drop January? The month the memory goes to mass production? Earliest will be march-may. Looks we we will have the same time line pascal. May be little early. But same timeline as pascal seems to be the thing here.
Click to expand...

I think both Micron and Samsung is way ahead of Hynix.

Also remember how GDDR5X couldn´t be possible? Yet GP104 got released with it.
 
CSI_PC said:
It also needs to be considered Pascal was a technical risk/milestone towards Volta as it introduced certain technologies with a node shrink while the rest will come with Volta or be further evolved (such as NVLink 1 to NVLink 2).
The way Pascal launched was anything but usual, because it is a forerunner for Volta.

GV100 is expected 15-16 months after GP100.
Early Q4 will mean 15-16 months between launch of GP104 and GV104.
Volta Xavier (Tegra) is on a similar or accelerated schedule (meaning the Volta architecture is well worked out and operational), this usually launches after both the initial Tesla and some consumer products with the same architecture.

TBH I thought it would be late Q4 for consumer (only for GV104 this year), but the option is there if Nvidia wants to be aggressive with this product initial cycle rollout across all segments earlier that the very latest rumours hint at albeit surmising the reason was consumer GPU sales down (we know Tesla-'Server' is through the roof in growth relative to consumer), and we know they can and do this even without competition from AMD (ala Pascal).
That said more recent analysis suggests sales for consumer GPU are still stable, so if Nvidia did move this forward I would think it relates to other reasons.
Cheers
Click to expand...

How would they release it in q4. If the memory for it won't be ready until 2018?
 
NKD said:
Really lower high end one? lol. So 980 goes to 1080 and a price bump. 1080 goes to 2080 and another price jump is justified? I don't remember things going down that road before 1080 release. 980 launched at 549, 1080 at 699. At what point price jump is too much? I guess never. Shit even iphones don't have this sort of pricing lol. Its just because of competition and nvidia knows no matter what fanboys will come buy and spend those dollars.
Click to expand...
The GTX 680 launched at $499, then the GTX 780 launched at $649.
 
NKD said:
You really expect it to drop January? The month the memory goes to mass production? Earliest will be march-may. Looks we we will have the same time line pascal. May be little early. But same timeline as pascal seems to be the thing here.
Click to expand...

NKD said:
How would they release it in q4. If the memory for it won't be ready until 2018?
Click to expand...

Yeah, you have two other companies that make gddr vram, and then you have sampling products well before mass production. So its possible to launch graphics cards as memory goes into mass production as long as there are no production problems. We also know nV has been helping Samsung with their GDDR5x development (higher clocked versions) so its not much of a far stretch for them to be also already pretty close to having GDDR6.

BTW AMD is also working on GDDR6 memory controllers already, so you think they can't do it?

https://www.linkedin.com/in/daehyun-jun-bb135a22/

- New DRAM controller : GDDR6
- Training algorithm
- Application
 
Armenius said:
The GTX 680 launched at $499, then the GTX 780 launched at $649.
Click to expand...

Ignoring the branding (which would otherwise be irresponsible), this is a Gx104 to Gx100 comparison; the GTX780 GPU is around twice the size and we can expect Nvidia to want to price it higher.

(and GTX780 will probably be the last time a Gx100 part makes it into a consumer [non-Titan] GPU until we see a paradigm shift in consumer computing, the extra compute is just meaningless for consumer workloads)
 
NKD said:
How would they release it in q4. If the memory for it won't be ready until 2018?
Click to expand...
Because GDDR6 is likely to be GV102 (sometime next year) and GDDR5X for GV104 (Q4).
For GV104 it is about diminishing returns, but maybe Tesla/Quadro will be GDDR6 (Nvidia comfortably split products between GDDR5 and GDDR5X on GP104 consumer even with the recent refresh)

The 12Gbps is now available for production orders from Micron in GDDR5X, and so far Nvidia is only using 11Gbps in the recent refresh so that is further future gains available.
But yeah at some point we will see GDDR5X become obsolete across all gpu tiers/products/segments.
Cheers
 
Last edited:
I'm jumping in on 1080ti(s) in sli. I doubt that a 2080 Ti will be around until 2019 so I'm not going to wait around any longer once the 1080ti FTW3's are in stock.
 
Armenius said:
The GTX 680 launched at $499, then the GTX 780 launched at $649.
Click to expand...

Also worth mentioning that:

The jump from GTX 580 to GTX 680 (if we are doing x80 to x80, ignoring die size changes between 580 and 680) is a mere 20%, if even that.

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_680/27.html

If you look at it from a die size point of view, you suddenly went from a $200 in a 560 to a $500 bracket in 680

If we completely and utterly ignored the fact that 680 even existed, and jump from 580 to 780 or 780ti (to keep die sizes the same)

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_780/26.html

780 is about 50~60% better for $649

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_780_Ti/27.html

780ti is about 85% better for $700 compared to 580

Now, let's us see the jump from 780 to 980 (again, keeping the x80 to x80, rather than die size)

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_980/26.html

A 33% increase for $550, making it a better deal than comparing the case from 580 to 680,

And we compare 780ti to 980ti (die size considerations, keeping the chip as fully enabled as possible)

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_980_Ti/31.html

980ti is about 40% better than 780ti for $650, possibly making 980ti the worst flagship to flagship upgrade, but 980ti was a mere architectural change to 780ti, whereas 780ti was both a change and node shrink against 580.

Finally, going from 980 to 1080

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_1080/26.html

80% increase of a 980 to 1080 for $599

980ti to 1080ti

https://www.techpowerup.com/reviews/NVIDIA/GeForce_GTX_1080_Ti/30.html

showing a similar 80% increase of a 980ti to 1080ti for $700

By looking at these comparisons, I find that Pascal's asking price as fairer, at least when compared to Kepler's asking prices. 1080ti costs the same to upgrade from a 980ti as 780ti was to 580, for similar performance gains, which makes pascal actually cheaper than kepler due to inflation. So no, I don't think Pascal is overcharging at all, especially considering that perf increase per dollar was much worse in Maxwell to Kepler.

Also, the upgrade from Fermi to Kepler is only comparable to Maxwell to Pascal if you ignore the entire 600 lineup existed, if we did, Kepler would, by far, have the worst perf increase to price ratio, far worse than Maxwell to Pascal.

EDIT: proof reading revealed several inaccuracies, corrected them

Adding a comment: another point with regards to 780ti vs 1080ti is that 780ti was in a worse position to ask for a $700 card than 1080ti was, given that AMD had just released 290x back before 780ti was released and was within about 10% in performance and also had more VRAM, AMD has no answer to 1080 now, let alone 1080ti, so from "milking" and "competition" perspective, I cannot help but to think that if 780ti wasn't considered overpriced, there is absolutely NO reason to consider 1080ti (or possibly even 1080 at its release non-FE MSRP) is overpriced, and this is analysing Kepler pretending 600 series didn't exist. FE edition card prices however, is a completely different story, but not so relevant now that 1080ti dropped FE pricing.

Also, I am quoting TPU exclusively due to their wider range of video cards tested during their reviews, hence it's easier to connect the dots between generations of GPUs.
 
Last edited:
All fine and dandy, but it doesn't detract from the fact the 680 was the high end in NVIDIA's stable for 14 months. If we ignore Kepler Gen1, does that also mean we ignore Fermi Gen1? Tesla Gen1?
 
Fudzilla: Nvidia to showcase Volta in early May

Volta, the successor to the Pascal architecture, will first debut as a data center, deep learning, artificial intelligence product, since margins in this market are much higher than with the gaming stuff.

...
One logical codename for the GP100 codename would be the GV100, as in Geforce Volta 100. A few magazines pointed out that Volta might come in 2018 with GDDR6, but Fudzilla expects that we will see Volta based GPUs probably before the end of summer, or should we say Q3 2017.

...We expect to see Volta using HBM 2 memory for the GV100 part and the GV104 - the GPU version of the chip is likely to use the GDDR5X with some Titan versions possible with HBM 2, but only if being really pushed by Vega.

Of course, Jensen will shortly share some official Volta information, in the second week of May.
Click to expand...

www.fudzilla.com/news/graphics/43485-nvidia-to-showcase-volta-in-early-may
 
So "high end" quoted by Hynix probably has the meaning of "Titan/x80ti" lineup, rather than also including x80/x70?
 
chenw said:
So "high end" quoted by Hynix probably has the meaning of "Titan/x80ti" lineup, rather than also including x80/x70?
Click to expand...

384bit=Certainly not GV104. GV100 with HBM2 and GV102 with 384bit from what it looks like.
 
I see no reason why we won't see Halo Volta in the summer of next year, it just matches up with nVidia's current cycling. I do doubt we will see Volta this year.

Also I don't think people using the word 'fanboy' understand what they are saying. NKD you cannot claim that nVidia buyers are fanboys for buying nVidia when AMD refuses to put out a halo product. I want my wine at date of purchase, not a year(s) after.

If AMD had a halo product the 'fanboy' argument would have more validity.
 
I still say it will be September '17 at the earliest, March '18 at the latest for the first consumer Volta products to hit the shelves.
 
It makes sense if they release 2050ti first like it was with maxwell to demonstrate a small piece of what the arch is capable of , but I see no reason to rush with mid and high end while amd is dragging its feet and won't offer anything more powerful than pascal.
 
brianj said:
I see.. I'm just affraid nvidia will release a new GPU soon
Click to expand...
I get it, and I'm in a similar boat.

It's past time for a new build in my case. But I teach school and will have about 2 months off this summer, at which time I'll get some overtime pay in and build a computer. Whether I buy parts to build in mid-May or late-July doesn't really matter to me. My 'needs' are actually 'wants.' So if there will be options available by August, I'll wait. If there won't be, I'll go ahead for May.

So yeah. Sometimes knowing release dates matters... because sometimes 'needs' are really 'wants' and can be waited for, but only to a point.

Right now planning on just nabbing a 1070, personally. But curious to see what else might be available in a similar price range by the time summer is through.
 
OrangeWolf said:
I get it, and I'm in a similar boat.

It's past time for a new build in my case. But I teach school and will have about 2 months off this summer, at which time I'll get some overtime pay in and build a computer. Whether I buy parts to build in mid-May or late-July doesn't really matter to me. My 'needs' are actually 'wants.' So if there will be options available by August, I'll wait. If there won't be, I'll go ahead for May.

So yeah. Sometimes knowing release dates matters... because sometimes 'needs' are really 'wants' and can be waited for, but only to a point.

Right now planning on just nabbing a 1070, personally. But curious to see what else might be available in a similar price range by the time summer is through.
Click to expand...

If nVidia releases per past history, next generation's mid-tier cards will be nipping at the heels of this generation's flagships. So, you could get a 'just enough to play what I want' or get a flagship and not worry until next generation's Ti comes out. IMO, it's always better to pick up a card in the early stages of a release so you can enjoy it for the entire length of that generation (not that this advice helps you at the moment).
 
Algrim said:
If nVidia releases per past history, next generation's mid-tier cards will be nipping at the heels of this generation's flagships. So, you could get a 'just enough to play what I want' or get a flagship and not worry until next generation's Ti comes out. IMO, it's always better to pick up a card in the early stages of a release so you can enjoy it for the entire length of that generation (not that this advice helps you at the moment).
Click to expand...

I would say the exception with that is the release of the 80ti of any generation, it's best to wait for aftermarket air cooled cards. With mid-range cards the ref coolers are quiet enough at 160-170W TDP, and with the Titan you can just throw on last gen's hybrid cooler at release and call it a day since there aren't going to be any other options later anyway, so the advice is solid.
 
Nvidia might surprise us and release new Ti/Titan parts that are compatible with existing waterblocks. To be honest the PCB for the Ti is very solid, I hope they don't change the layout.

And then you'll only have to buy a water block once :cry:
 
Oh so you guys are ready to buy the GV104 (2080) part for $650?
Even if Volta comes this year, pass on it. Within a few months faster parts will be made.
In the meantime the 1080Ti is a kickass card and that's available right now.

My vote is for March 2018. But don't buy the top card. The GTX 2080 or a cut Titan card are bad investments.
 
WorldExclusive said:
Oh so you guys are ready to buy the GV104 (2080) part for $650?
Even if Volta comes this year, pass on it. Within a few months faster parts will be made.
In the meantime the 1080Ti is a kickass card and that's available right now.

My vote is for March 2018. But don't buy the top card. The GTX 2080 or a cut Titan card are bad investments.
Click to expand...
not a bad investment if you plat at 1440P 120hz or 4K 60hz or VR
 
WorldExclusive said:
Oh so you guys are ready to buy the GV104 (2080) part for $650?
Even if Volta comes this year, pass on it. Within a few months faster parts will be made.
In the meantime the 1080Ti is a kickass card and that's available right now.

My vote is for March 2018. But don't buy the top card. The GTX 2080 or a cut Titan card are bad investments.
Click to expand...

SomeGuy133 said:
not a bad investment if you plat at 1440P 120hz or 4K 60hz or VR
Click to expand...

You guys should join us here. https://hardforum.com/threads/i-jus...ext-gpu-whos-with-me.1930921/#post-1042969874.
 
Algrim said:
If nVidia releases per past history, next generation's mid-tier cards will be nipping at the heels of this generation's flagships. So, you could get a 'just enough to play what I want' or get a flagship and not worry until next generation's Ti comes out. IMO, it's always better to pick up a card in the early stages of a release so you can enjoy it for the entire length of that generation (not that this advice helps you at the moment).
Click to expand...

This.
But that also means if Vega is around 1080-1080Ti and closer to the Ti, then it's still competitive for a while.
Problem is neither is still not really enough for 4k, especially with higher refresh rate screens coming.
What many forget to realize is screen resolution and refresh rate is increasing faster than GPUs or even single cables can handle at the moment.
 
N4CR said:
This.
But that also means if Vega is around 1080-1080Ti and closer to the Ti, then it's still competitive for a while.
Problem is neither is still not really enough for 4k, especially with higher refresh rate screens coming.
What many forget to realize is screen resolution and refresh rate is increasing faster than GPUs or even single cables can handle at the moment.
Click to expand...

And I hope Vega is competitive (though I am not in the market for a new card). We need a healthy AMD/RTG.
 
You must log in or register to reply here.
Back
Top