Nvidia Volta 20 series coming Q3 2017

Griffildur · 23 Apr 2017 at 17:16

TNA said:
Yea, but wccftech is a very good source and is deemed quote worthy in these parts

sarcasm meter high ...

TNA · 23 Apr 2017 at 17:17

Griffildur said:
sarcasm meter high ...

Mtom · 23 Apr 2017 at 17:19

I'd bet these will not show up until 2H 2018

TNA · 23 Apr 2017 at 17:24

Mtom said:
I'd bet these will not show up until 2H 2018

I bet they will

varkanoid · 23 Apr 2017 at 17:36

I was being rightatchya but it was a bit of a fail. I think the new forum should have come with new smileys.

valerislysander · 23 Apr 2017 at 19:04

I think Volta for gaming will be 1st quarter 2018. That seems to fit with nvidias release cycle. As with pascal,enterprise products first probably 4th quarter this year.

I really don't think AMDs Vega will outperform the 1080 or 1070 maybe match them for slightly cheaper price seeing as the nvidias have dropped a lot.

TNA · 23 Apr 2017 at 19:10

valerislysander said:
I think Volta for gaming will be 1st quarter 2018. That seems to fit with nvidias release cycle. As with pascal,enterprise products first probably 4th quarter this year.

I really don't think AMDs Vega will outperform the 1080 or 1070 maybe match them for slightly cheaper price seeing as the nvidias have dropped a lot.

No change of that imo. Seems like more an emotional conclusion than logical. If Vega cannot beat the 1080, I will eat my hat. Lol

thenoseknows · 23 Apr 2017 at 19:16

TNA said:
You do realise I was being sarcastic right?

See this is why I feel the need to use /s at times

A sarcasm detector would be real useful.

Samwell · 23 Apr 2017 at 19:30

So guys, how about a really believable news about Volta and not this made up story?

SK Hynix Inc. (or ‘the Company’, www.skhynix.com) today introduced the world’s fastest 2Znm 8Gb(Gigabit) GDDR6(Graphics DDR6) DRAM. The product operates with an I/O data rate of 16Gbps(Gigabits per second) per pin, which is the industry’s fastest. With a forthcoming high-end graphics card of 384-bit I/Os, this DRAM processes up to 768GB(Gigabytes) of graphics data per second. SK Hynix has been planning to mass produce the product for a client to release high-end graphics card by early 2018 equipped with high performance GDDR6 DRAMs.

http://www.skhynix.com/eng/pr/pressReleaseView.do?seq=2086&offset=1
As AMD is using HBM i bet this 384Bit GPU is Nvidia next high end coming in early 2018, so 16 GBps with 384 to a nice 768 GB/s Bandwidth sounds already good.

D.P. · 23 Apr 2017 at 19:38

Samwell said:
So guys, how about a really believable news about Volta and not this made up story?

http://www.skhynix.com/eng/pr/pressReleaseView.do?seq=2086&offset=1
As AMD is using HBM i bet this 384Bit GPU is Nvidia next high end coming in early 2018, so 16 GBps with 384 to a nice 768 GB/s Bandwidth sounds already good.

About as definitive as it comes really. Onlyraly question is will the GDDR6 appear on high end Volta only, wiith GDDR5x on the mid-range? 2080 with 5x rleased Q42 or Q4 2017, TitanV and Ti with 6 released Q1/2 2018?

Samwell · 23 Apr 2017 at 19:49

I think it will both need GDDR6. This gpu will have nearly 40% more bandwidth than GP102, a GP104 will have a 256Bit interface and will also need the same amount more to get fast enough, because else the difference will be too big. This news sounds like Big Chip first for me, so Q1 2018 Titan and Q2 maybe 2080Ti and the GV104 stuff. I don't believe in 2017 Volta for consumers.

Might be that Nvidia is expecting big stuff from vega and so they get the big guns out first to get on top again as fast as possible.

TNA · 23 Apr 2017 at 20:07

TNA said:
No change of that imo. Seems like more an emotional conclusion than logical. If Vega cannot beat the 1080, I will eat my hat. Lol

Hard to detect your sarcasm it seems. No emojis. What he said could easily be serious

AthlonXP1800 · 23 Apr 2017 at 20:20

Samwell said:
So guys, how about a really believable news about Volta and not this made up story?

http://www.skhynix.com/eng/pr/pressReleaseView.do?seq=2086&offset=1
As AMD is using HBM i bet this 384Bit GPU is Nvidia next high end coming in early 2018, so 16 GBps with 384 to a nice 768 GB/s Bandwidth sounds already good.

Nvidia will use Hynix memory for high end Volta GPU. Interesting.

I think it easy to guess what GPU it will be use 384 bit. That will be GTX 2080 Ti.

Nvidia will launch GTX 2080 Ti a year after GTX 1080 Ti.

Maybe Nvidia will launch flagship Volta Titan X with HBM2 memory in August 2017, exactly a year after Pascal Titan X launch back in 2016 then a month later in September would see Nvidia launch GTX 2070 and GTX 2080 with 256 bit GDDR6 Micron memory.

JediFragger · 23 Apr 2017 at 20:55

AthlonXP1800 said:
Maybe Nvidia will launch flagship Volta Titan X with HBM2 memory in August 2017, exactly a year after Pascal Titan X launch back in 2016 then a month later in September would see Nvidia launch GTX 2070 and GTX 2080 with 256 bit GDDR6 Micron memory.

I can see that as vice-versa tbh, with 2070/2080 coming in Sept and the Titan coming in Jan/Feb. Seems to fit their previous M.O.

Greebo · 24 Apr 2017 at 09:37

jjgreenwood said:
1080 was 25-35% quicker than 980ti no reason why 2080 won't repeat that

And maybe more as isnt Volta touted to be a major step up, much bigger than the jump to Pascal? Perhaps Nvidia can then release the 2070 first and keep the 1080 pricing?

nashathedog · 24 Apr 2017 at 09:58

NVIDIA Volta with GDDR6 in early 2018?

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

It seems to me that Nvidia really don't want to use HBM in their gaming cards,

I still think it's because of the way Nvidia's software overclocks the core on the fly and the HBM being on the same die is the problem.
I reckon that's why the Fiji cards are such poor overclockers and I'll be surprised if Vega is any better.
Maybe AMD will find a way around the issue in time for Vega or Navi,
Once they do (if they do) Nvidia can get one of AMD's cards and
nick the solution but for now Nvidia seem more interested in improved GDDR solutions.

JediFragger · 24 Apr 2017 at 10:01

nashathedog said:
NVIDIA Volta with GDDR6 in early 2018?

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

It seems to me that Nvidia really don't want to use HBM in their gaming cards,

I still think it's because of the way Nvidia's software overclocks the core on the fly and the HBM being on the same die is the problem.
I reckon that's why the Fiji cards are such poor overclockers and I'll be surprised if Vega is any better.
Maybe they'll have found a way around the issue with Vega or Navi and once they do (if they do) Nvidia can get one of AMD's cards and
nick the solution but for now they seem more interested in improved GDDR solutions.

I see what you're getting at but my Fury went up and down on the core no problem

I personally think it's more of a cost issue, and the fact that their modelling has probably worked out that any potential gain by using HBM isn't worth it...

nashathedog · 24 Apr 2017 at 10:30

JediFragger said:
I see what you're getting at but my Fury went up and down on the core no problem

I personally think it's more of a cost issue, and the fact that their modelling has probably worked out that any potential gain by using HBM isn't worth it...

You may be right, It's as plausable as my theory, maybe more so. I think that HBM does not like heat and does not handle large heat variations very well which is why The X has watercooling and all the Pro's have such big coolers on them, As for the Nano's, they avoid the problem by using carefully selected chips and reducing the power available to it. Nvidia's GPU boost pushes the clocks up and down a lot with temps following suit and that's what I think is HBM's Achilles heal.
It's only a layman's theory though.

TNA · 24 Apr 2017 at 10:41

nashathedog said:
You may be right, It's as plausable as my theory, maybe more so. I think that HBM does not like heat and does not handle large heat variations very well which is why The X has watercooling and all the Pro's have such big coolers on them, As for the Nano's, they avoid the problem by using carefully selected chips and reducing the power available to it. Nvidia's GPU boost pushes the clocks up and down a lot with temps following suit and that's what I think is HBM's Achilles heal.
It's only a layman's theory though.

Could be right. HBM might need more time to mature. I am happy we are finally seeeing GDDR6, at least we will have the option of either that or HBM2. Options are always good and GGDR6 will hopefully mean cards coming out making use of the extra bandwidth by design.

D.P. · 24 Apr 2017 at 12:21

nashathedog said:
NVIDIA Volta with GDDR6 in early 2018?

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

It seems to me that Nvidia really don't want to use HBM in their gaming cards,

I still think it's because of the way Nvidia's software overclocks the core on the fly and the HBM being on the same die is the problem.
I reckon that's why the Fiji cards are such poor overclockers and I'll be surprised if Vega is any better.
Maybe AMD will find a way around the issue in time for Vega or Navi,
Once they do (if they do) Nvidia can get one of AMD's cards and
nick the solution but for now Nvidia seem more interested in improved GDDR solutions.

It is not the same die, so no, totally wrong.

Nvidia likely don't want to use HBM2 because supply is likely still low, prices high, the interposer expensive and yields are reduced and at the end if the day downstairs yet bro g significant performance enhancements to gaming. Nvidia are happy to use HBM2 where profit margins are high, volume low and high BW memory makes a big difference - HPC and compute