NVIDIA Posts Cryptic #BeForTheGame Video Pointing at 20th August

[cucker tarlson]

I'm pretty certain be for the game is correct syntactically.

 

[rtwjunkie]

Why would they release a flagship that's slower than 1080Ti from march 17 ? We don't know the clock yet,though 750mm2 turing has a higher boost clock than P100 (1.75Ghz vs 1.44Ghz, +20%). There's also improvements to SM and CUDA themselves. I don't know hwo you calculated 2080 performance,but I'm 99,99% sure you're wrong.

Because 2080 is not a flagship. Nothing built on an x04 is ever the line’s flagship. By the very nature of the chip it occupies a tier below.

 

[cucker tarlson]

Because 2080 is not a flagship. Nothing built on an x04 is ever the line’s flagship. By the very nature of the chip it occupies a tier below.

980 beat 780Ti by, I don't remember, 10% or something. 1080 beat 980Ti by +20%. They release the 104 chips first cause they can beat the last gen 100/200.

 

[rtwjunkie]

980 beat 780Ti by, I don't remember, 10% or somthing. 1080 beat 980Ti by +20%.

Now you’re being sarcastic. They were both faster. They weren’t the flagship of either gen though. They are built on the GM104 and GP104 chips, which occupy the high-middle range of Nvidia’s chip tiers.

There has always been a one step down improvement in next line over previous lineup models.

This is why you don’t buy down either, just because it is faster than your current card once you are at a level (I.e. 1080ti to 2080). Always buy the same tier or higher as a replacement, or eventually you will end up with a x030 model as your primary card, wondering why you can only play your old games at full throttle.

 

[IceShroom]

Do you know why people go to SIGGRAPH?

Yeah, to see New GPU.

 

[cucker tarlson]

That's why I wonder why ppn claims something completely different. I know all RT features are fancy, but if the card doesn't deliver in overall performance it's going to be a flop.

what you are claiming with tiers is more dependent on the pace at which gaming industry progresses and your target resolution than some sort of formula. my 1080 is doing better two years after launch than my 980 did two years after launch, cause 2014-2016 saw a lot of very demanding games while 2017-18 was pretty stagnant, meanwhile I stayed at the same resolution. Next two years will be stagnation too except for this ray tracing stuff until next gen consoles launch, but that's like 2 years from now.

 

[Prince Valiant]

That's why I wonder why ppn claims something completely different. I know all RT features are fancy, but if the card doesn't deliver in overall performance it's going to be a flop.

what you are claiming with tiers is more dependent on the pace at which gaming industry progresses and your target resolution than some sort of formula. my 1080 is doing better two years after launch than my 980 did two years after launch, cause 2014-2016 saw a lot of very demanding games while 2017-18 was pretty stagnant, meanwhile I stayed at the same resolution. Next two years will be stagnation too except for this ray tracing stuff until next gen consoles launch, but that's like 2 years from now.

I disagree, Nvidia has built market share on the premise of features like RT/Shadowplay/etc.

 

[cucker tarlson]

No,it has built it by knowing how to develop and release their products in such a way as to always top the performance charts. When did you last see any TPU performance chart with AMD card as #1 ?

 

[ppn]

That's why I wonder why ....

2080 was supposed to be on 7nm @2.5Ghz clocks, not 12nm @ 1.75Ghz, hence we won't see 2080 being faster than 1080Ti the way 1080 was to 980Ti. it is not a Shrink yet.

See, 500mm2 is insanely big for a 256 bit memory type. usually in the 260-330 mm2 range. RTX 3080 on 7nm will fix things.

 

[cucker tarlson]

Supposed by whom ? I have never seen anyone in a sane mind claim next gen is 7nm and 2.5GHz.
500mm2 is the whole chip with rt and tensor, not a traditional fp32 -only one. 980 was 400mm2 so not all of them are 260-300mm2.

Now you're saying "2080 will not be faster than 1080Ti like 1080 was to 980Ti", before you said it will be slower. I'm pretty sure you're trolling in most of your statements.

 

[Vayra86]

I have always had a very solid bet on a repeat of the kepler refresh and that may well be happening here after all. Small perf bumps at the same price point and some new useless tech to make it seem like you need this.

Its economically the best move for Nvidia. Still is. And I dont see how they will make huge jumps from Pascal either especially when they are adding a new resource to the die. Full GV100 per core performance is too low to make the jumps some ppl claim.

 

[cucker tarlson]

I have always had a very solid bet on a repeat of the kepler refresh and that may well be happening here after all. Small perf bumps at the same price point and some new useless tech to make it seem like you need this.

Its economically the best move for Nvidia. Still is. And I dont see how they will make huge jumps from Pascal either especially when they are adding a new resource to the die.

That was a pretty safe bet to see sometging similar to Kepler, though 780 was still 20% faster than 680. This time we'll see 2080 no more than 10% faster than 1080Ti, packed with RT features.
If it can't beat 1080Ti on general performance it's going to be a flop, despite RT features.

 

[Vayra86]

That was a pretty safe bet to see sometging similar to Kepler, though 780 was still 20% faster than 680. This time we'll see 2080 no more than 10% faster than 1080Ti, packed with RT features.

If 2080 is 10% faster then it is not a repeat of that because it would add +50% at the same tier instead of 20. Whats different with Kepler is that they didnt have or needed an x100/110 SKU they could transplant to the refresh but rather introduced it to the Geforce stack.

 

[B-Real]

After Fermi, xx80 has always been full xx-104 chip. I don't see reason why would they change that. I would expect 2080 to be quite equal with gtx1080ti in performance. But that all depends on achievable clocks(Some says Turing can overclock as high as 2.5GHz).



It has new tech what old flagship does not have. And because amd has no competitive highest end product, nvidia could add some new features instead of going brute force with core count. It is plausible that there were smaller Volta's on the table and turing would have been 7nm product. But without real competition they went with new tech and old very well known 12nm(16nm FF++) manufacturing process instead.

Which would be a big step-back after Pascal and even Maxwell. GTX 980 was about 10% faster than the 780Ti and the 1080 was about 35% faster than the 980Ti. So a 1180 (or 2080, whatever) being equal with the 1080Ti would be a disappointment for most gamers looking for high-end GPUs, regardless of the next AMD GPU's performance.

 

[cucker tarlson]

Lol no way it was 35%, 25% at max, ~20% on average.

 

[StrayKAT]

Is this another odd Taiwanese slogan, like ASUS "In Search of Incredible"?

They deserve to be ignored (from me) just for this alone. lol

 

[dj-electric]

Yeah, to see New GPU.

Did the engine developers, and 3D artists, and professional modelers see a new GPU at SIGGRAPH this year?

 

[IceShroom]

Did the engine developers, and 3D artists, and professional modelers see a new GPU at SIGGRAPH this year?

Yeah. But isn't Nvidia for gamers first and other last?

 

[StrayKAT]

Yeah. But isn't Nvidia for gamers first and other last?

Well, they're easily winning the gamer market. The other not so much.. so I can see why they're focused on it.

 

[IceShroom]

Well, they're easily winning the gamer market. The other not so much.. so I can see why they're focused on it.

They are winning doesen't mean they need not release new GPUs? Well Intel released several good comsumer class CPU during Bulldozer time. If Nvidia for gamers they need to release new even there is no competitor.

 

[StrayKAT]

They are winning doesen't mean they need not release new GPUs? Well Intel released several good comsumer class CPU during Bulldozer time. If Nvidia for gamers they need to release new even there is no competitor.

Oh I misread that. I didn't mean not releasing GPUs. I mean that they might want to focus on and invite workstation users at this conference for now?

 

[dj-electric]

Yeah. But isn't Nvidia for gamers first and other last?

What kinds of things do gamers do? What type of products do they consume?

 

[rtwjunkie]

Nvidia is focused far beyond gaming. In fact I would be surprised if gaming GPU’s are even 40% of their market.

 

[cucker tarlson]

Nvidia is focused far beyond gaming. In fact I would be surprised if gaming GPU’s are even 40% of their market.

 

[StrayKAT]

In other news, I wonder how much that auto segment is getting hit because of Tesla

Then again, I guess Tesla isn't all that big.