• Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.

NVIDIA Volta Unveiled (GV100 new GPU Architecture)

joesiv

Member
That's weird, I would have thought that would be a basic thing. Does AMD have the same problems or do they not create workstation GPUs?
Yes, a lot of the "driver" for consumer graphics cards are optimizations for specific games, and such. This types of cards are focused on compute. Other workstation cards, like Nvidia Quadro or AMD FireGL, are very similar to their consumer gaming cards, but the driver will lack those same gaming specific optmizations. They support DirectX and Open GL of course, but you will get less performance on them than a gaming card, despite the specs being comparable.

I don't know about these volta deep learning cards though, but with Quadro cards you could driver hack them into believing they're a GeForce card (and visa versa on some models of geForce cards)
 

Kaako

Felium Defensor
Can we get a super fast GPU for your next enthusiast consumer product please, Nvidia? At least 2.5GHz+ core clock factory default w/out OC? With insane fast memory & BW of course. You're gonna charge an arm and a leg regardless, so hook us up next round.
 

Tryxx

Member
Every single time I talk myself into finally upgrading my 970 something like this comes out and then I want to wait lol. I hate this!
 

Newboi

Member
How Much weaker is generally the game version compared to Business version?

You're going to have to clarify what you're asking. The gaming card will generally be better at gaming than the business variant. Mostly due to clock speed and driver enhancements usually.

The business variant, especially in the case of Volta, will contain specialized units and better FP64 performance that's more suited to HPC and workstation oriented projects. Also the drivers for those cards are certified, better tested, and better optimized for business use cases.
 

Eylos

Banned
You're going to have to clarify what you're asking. The gaming card will generally be better at gaming than the business variant. Mostly due to clock speed and driver enhancements usually.

The business variant, especially in the case of Volta, will contain specialized units and better FP64 performance that's more suited to HPC and workstation oriented projects. Also the drivers for those cards are certified, better tested, and better optimized for business use cases.
Got It, so the gaming card Will do 15 tflops too? Thats Very incredible
 
That 970 I intended to sell after upgrading last year is quickly going cost more to ship than the price I'll get for it huh?

Coming soon to the Free to a Good Home thread....
 

Kambing

Member
Guys when it comes to next gen consoles, well the PS5 in this case, assuming it launches in 2020, would these cards (the eventual 2080) be in the ballpark of performance? 15 Tflop consoles in-coming?

I've got a 1080ti and really happy with its 4k performance.
 

Newboi

Member
Any chance the 2070 and 2080 will be cheaper than the 1070 and 1080 at launch?

I highly highly doubt this. Nvidia tested the market with overpriced founders edition cards and consumers ate them up. You'll also have to deal with OEMs jacking up prices due to consumer demand.

Competition from AMD might bring down prices some, but it's hard to believe AMD will be truly competitive at the high-end for a while. There's also the fact that Nvidia is effectively giving AMD room to charge more than they would normally too.
 
Guys when it comes to next gen consoles, well the PS5 in this case, assuming it launches in 2020, would these cards (the eventual 2080) be in the ballpark of performance? 15 Tflop consoles in-coming?

I've got a 1080ti and really happy with its 4k performance.

These cards will be ahead of consoles (2080ti will be for sure, 2080 OCed should be), your 1080ti gets around 14Tflops if clocked to 2Ghz. I do think 15Tflop consoles will be possible. According to Anandtech that transition should give a 60% power reduction and a 70% area reduction. A Scorpio like machine could scale to around 15Tflops just using the density and power advantages of the node.
 

E-Cat

Member
According to Anandtech that transition should give a 60% power reduction and a 70% area reduction. A Scorpio like machine could scale to around 15Tflops just using the density and power advantages of the node.
16nm-->7nm is going to be the greatest litho process leap ever; even greater than 28nm-->16nm, which had a 50% area reduction. Power reduction was slightly more at 70% thanks to the introduction of FinFET, though.
 
My general rule of thumb is that if there is a card available that fulfills the needs you have, go ahead an buy it. There will always be a next best thing, especially when it comes to GPUs.
Replace 'card' with 'piece of tech' and you have pretty solid advice in general. Hype trains are fun, but they're pointless if you never get off. The exception is generally with things that come one or twice a decade, such as brand new cpu archs (I couldn't fault people waiting for Ryzen even if Intel CPUs were just fine).
 
Replace 'card' with 'piece of tech' and you have pretty solid advice in general. Hype trains are fun, but they're pointless if you never get off. The exception is generally with things that come one or twice a decade, such as brand new cpu archs (I couldn't fault people waiting for Ryzen even if Intel CPUs were just fine).

Agreed fully.
 

Nerrel

Member
Any chance the 2070 and 2080 will be cheaper than the 1070 and 1080 at launch?

Probably not. Worst case scenario, I'm hoping a $200 2060 turns out to be a worthwhile upgrade from a 970. I assume it'll just outperform the 1070 and be great for 1080p.
 

ethomaz

Banned
Actually Titan Xp with modest Overclock to 2ghz does over 15Tf so i' m expecting new Titan/Ti VOlta overclocked to 2ghz be over 20-22TF or even more
Depend on clock but new Titan will be based on GV102 that probably will have 5376 SPs.

1700Mhz = ~18.2TFs
1800Mhz = ~19.3TFs
1900Mhz = ~20.4TFs
2000Mhz = ~21.5TFs
 

ZOONAMI

Junior Member
It just feels like NV has AMD completely out matched in the GPU space.

At least AMD largely managed to catch Intel but it might all be too little too late.

I say this as an AMD shareholder and fan.
 

ZOONAMI

Junior Member
Probably not. Worst case scenario, I'm hoping a $200 2060 turns out to be a worthwhile upgrade from a 970. I assume it'll just outperform the 1070 and be great for 1080p.

Did they confirm they are skipping 11 series naming??
 
Realistically:

2060 - $349
2070 - $459
2080 - $749
Titan V - $1400
2080 Ti - $799 (consider this will release several months after the 2080. The 1080 and 1080 Ti actually had the same $699 launch price)
 

Reallink

Member
Any chance the 2070 and 2080 will be cheaper than the 1070 and 1080 at launch?

There's more chance of Israel forming a military alliance with Iran.

It's unlikely but not completely impossible, depends on the yields of the new 12nm node and AMD's competing offerings. The recent price drop and cheaper than expected Ti were reputedly due to the sales numbers not being where they wanted them, and based on Steam's HW survey, the much cheaper 1060 has cannibalized the overpriced 1070 (that has only started seeing gains as street prices have plummeted in recent months). Meanwhile the 970 just completely annihilated everything else in sales volume in it's day (including the 960). The first edition tax on the 1070 and ridiculous $450 street prices was absolutely a mistake, they would have easily done double the volume around $350.
 

Renekton

Member
It's unlikely but not completely impossible, depends on the yields of the new 12nm node and AMD's competing offerings. The recent price drop and cheaper than expected Ti were reputedly due to the sales numbers not being where they wanted them, and based on Steam's HW survey, the much cheaper 1060 has cannibalized the overpriced 1070 (that has only started seeing gains as street prices have plummeted in recent months). Meanwhile the 970 just completely annihilated everything else in sales volume in it's day (including the 960). The first edition tax on the 1070 and ridiculous $450 street prices was absolutely a mistake, they would have easily done double the volume around $350.
The 12nm node is just a refinement on existing 16FF+ so yields could be fairly good right off the bat.
 

rrs

Member
It's unlikely but not completely impossible, depends on the yields of the new 12nm node and AMD's competing offerings. The recent price drop and cheaper than expected Ti were reputedly due to the sales numbers not being where they wanted them, and based on Steam's HW survey, the much cheaper 1060 has cannibalized the overpriced 1070 (that has only started seeing gains as street prices have plummeted in recent months). Meanwhile the 970 just completely annihilated everything else in sales volume in it's day (including the 960). The first edition tax on the 1070 and ridiculous $450 street prices was absolutely a mistake, they would have easily done double the volume around $350.
the 970 was a dark era where the $200 GPU market was dead because bitcoin fucked AMD's gaming marketshare and the 970 was the best value card. You could see nvidia's smugness in real time as the pack-in games went from the fucking witcher 3 and MGSV to a ubisoft title
 

dr_rus

Member
You guys do know 11 comes after 10?
I'm not a fan of either 11xx or 20xx and hope that NV will use the disruption of architecture change to restart the numbering with some suffix. Like GeForce Vx0 (V60, V70, V80, etc) or something.

People always expecting a big price jump, but time and time again it never happens.
"People" generally don't. Some very specific part of people do.
 
I'm not a fan of either 11xx or 20xx and hope that NV will use the disruption of architecture change to restart the numbering with some suffix. Like GeForce Vx0 (V60, V70, V80, etc) or something.

They're never going to stick with a consistent scheme, tech companies are allergic to it.

They're never going to include year in there because then if they release in december or don't refresh for a while they start to look bad. They're hesitant to include the architecture name because there's no clear way for consumers to tell which one is better than another (Is Hawaii better than Fiji? Is Maxwell better than Kepler???), and if there was, people would get annoyed by rebadges and low end cards using old archs. They don't want to stick with a consistent numbering scheme because they run out fast and have to compete with competitor numbering schemes that may "look" superficially better.

IMO since consumer GPUs are an enthusiast market it wouldn't matter for any of those reasons, but I think their marketing believes that the ignorant people are a sufficiently large chunk of their market that they probably want to keep trucking with the "let's just fucking throw shit at the wall and maximize confusion" scheme they all use.
 
Top Bottom