D
Deleted member 13524
Guest
My use of implementation referred to current hardware+software implementation. It's still early know if the BVH units are a decent hardware implementation for real-time raytracing in videogames.
Nvidia Turing Product Reviews and Previews: (Super, TI, 2080, 2070, 2060, 1660, etc)
- Thread starter Ike Turner
- Start date
Because thats what you want to read/hear, clear message is that RTX offers a substantial rasterization improvement with very good RT performance for being first with it, just the Quake 2 RT shows enough.
In a time with diminishing returns i think RTX 20xx offers a very nice normal rasterization improvement, the addition of RT, DLSS, VRS, Mesh Shading etc only future proofed it.
With over 6 million sold it aint too bad. RT enabled made it the ultimate version to experience, too.
It may not or it may go well, pc version will offer the best visual experience, ppl that can opt for that version.
D
Deleted member 2197
Guest
7.3 million sold, slightly less than their target of 8.3 million which is extremely good.
Really?? Doesn't the GTX 1080Ti beat the 2070 at a lower price point? If there's any improvement in rasterisation performance (not features, of course) then I'd describe it as marginal, not substantial. It's only substantial in a 'money's no object, here's a bigger slab of silicon than every before which of course is going to be faster." And when it comes to architecture, the same die size dedicated to 1080 level architecture would apparently have produced a faster GPU for current and near-future games then RTX. RTX should get better performance when modern features are implemented in games.
D
Deleted member 13524
Guest
I want to hear that nvidia sold substantially less RTX cards than they predicted, as told on the record by their own CEO?
That adoption compared to Pascal is substantially slower, even when comparing cards of similar prices?
Am I really the one here with selective hearing?
Clear message is it does not.
Not on performance/die area nor on performance/dollar.
The TU106 is almost 50% larger than a GP104 and performs about the same. TU104 is 15% larger than GP102 and performs about the same.
Power consumption is very similar among the two pairs, despite RTX using the much newer 12FFN process.
Number of sales on a multiplayer game doesn't tell you if the game is being played by a lot of people.
EA closed down the ability to see online player counts after people started noticing that Battlefront 2 (with 9 million sales) was getting player counts lower than 2013's Battlefield 4 (which sold 7 million).
You know Battlefield V's 7 million sales are very short on expectations when you consider that its predecessor reached almost 3x more, and on the fact that EA had to lower their earnings forecast due to that game.
Silent_Buddha
Legend
Steam wasn't the only ones hurt by this. GoG and Origin are hurt by this as well. Both storefronts while not taking pre-orders unlike Steam, do sell the entire Metro series of games. Now they can't sell Metro: Exodus either because Epic paid big bucks to make Metro: Exodus exclusive to their storefront, giving consumers the big middle finger. DRM (denuvo) meant that a GoG release was always going to be delayed, but it likely would have shown up on Origin rather quickly.
What they should have done was just have Metro: Exodus on their storefront at 10-20 USD cheaper than other storefronts. Fostering competition and allowing the consumer to choose where they want to buy it. Instead they'd rather use their cash to buy exclusivity and limit consumer choice.
That's why there's an increasing call on the internet to boycott this game and the Epic store. People who are against piracy even going as far as to encourage people to pirate the game.
Regards,
SB
Exactly, 1080 was the pascal flagship for a while, 1080TI replaced that. The mid range 2070 performing like a high-end 1080TI is outstanding achievement.
I know what you want to hear, it doesnt affect Turing performance though.
Its very clear it does, 2080Ti is a whole lot faster then 1080Ti, so is the 2080<1080, 2070<1070 and 2060<1060. That is in normal rasterization, in a time where we cant expect big jumps in performance/diminishing returns. Thats why companies like nvidia have to come with new tech like RT, VRS, DLSS etc
Thats still over 6 million sold, your speculating that none of these 6 million even played it. Comparing it to your sony platform, some over 7 million Horizon Zero Dawn was sold, how many still play it i dont know, but not many seeing its a single player game. If you say not many got the BFV tech then not many did either HZD which has been out for much much longer.
D
Deleted member 2197
Guest
DuckThor Evil
Legend
Where have you seen 1080Tis being sold cheaper than the 2070? I see 2070s going for 500$/€, I have never seen a 1080Ti going for that price.
Typically Ebay has several 11GB 1080 TIs listed with "buy it now" for around $550. Heres one of them: https://www.ebay.com/itm/DELL-GTX-1...0VBcXPDx:sc:USPSPriority!44134!US!-1&LH_BIN=1 sometimes you can find people selling them for cheaper.
I stand corrected. A quick search suggested the 1080 was cheap probably because I was getting bunk results and not paying attention - actually it's hard to come by. 2070 achieves better results with less transistors, meaning regardless of market prices (affected by profit margins and RAM prices), 2070 is a cheaper device to manufacture while running faster than GTX 1080 TI, and getting similar framerates (I'm seeing some benchmarks - very quick Google - where 2070 is lower, and others where it's higher) with less shader cores and similar clocks. So a notably improved architecture.
Last edited:
2070 is comparable to a 1080 not to a 1080 Ti
2070 definitively a lot slower than a 1080 Ti. (I have both, and I don't play BFV on the 2070)
144 TMUs / 2304 shaders are no match for 224 TMUs / 3584 shaders
The Turing SMs are more efficient than the Pascal SMs, but not by 55%
In terms of die size Turing offers significant less bang for die size:
445 mm2 for 2070 compared to 471 for 1080 Ti (and 314 mm for 1080)
Last edited:
A1xLLcqAgt0qc2RyMz0y
Veteran
That is a USED unit. Seller refurbished means they blew out the dust and tested the unit. There is no warranty beyond the eBay's 30 day.
When prices are discussed like the claim that GTX 1080 Ti's have been priced lower that RTX 2070's it should be about NEW units with a factory warranty.
Just like in one of my pc's i have a i7 920, even at stock its no problem to run most titles. Hardware jumps arent that huge anymore, thats why i think its important we move to new tech.
Halo parts tend to last a lot longer than 2 years though.
It looks to be doing just that by being very close to 2080 and radeon vii in traditional gaming performance. 1080 being year older is probably aging really nicely as well.
VideoCardz just revealed chip photos of TU116 aka GTX 1660 Ti, that should about wrap it up as I don't think there's anything that hasn't leaked yet (except for each individual manufacturers all models ofc but you get the point)
https://videocardz.com/80034/msi-geforce-gtx-1660-ti-ventus-xs-and-tu116-gpu-pictured
My quick estimation for die size comes around 290mm^2 or slightly under
edit: also, the MSI GTX 1660 Ti Ventus XS PCB is built for 256-bit membus, which obviously isn't used in 1660 Ti - is the chip pincompatible with something that has 256-bit membus or does the chip really feature 256-bit, just partly disabled?
https://videocardz.com/80034/msi-geforce-gtx-1660-ti-ventus-xs-and-tu116-gpu-pictured
My quick estimation for die size comes around 290mm^2 or slightly under
edit: also, the MSI GTX 1660 Ti Ventus XS PCB is built for 256-bit membus, which obviously isn't used in 1660 Ti - is the chip pincompatible with something that has 256-bit membus or does the chip really feature 256-bit, just partly disabled?
Last edited:
D
Deleted member 13524
Guest
This card's PCB shows 2 unpopulated spaces for two extra GDDR6 chips.
Could a GTX 1670/1680 be coming, equipped with a 256bit bus?
The chip is almost as big as a GP104. Maybe there are some unused SM units in there too.
EDIT: Maybe not. The first GTX 1060 PCBs also had two unused GDDR5 spaces.
iMacmatician
Regular
My die size estimate is ~287 mm^2 with a range of [272 mm^2, 305 mm^2].
That's a lot bigger than any xx6 die from Fermi to Pascal, and similar in size to the 294 mm^2 GK104.
If this becomes a trend, then maybe Volta and Turing have ushered in a lineup-wide die size increase similar to the increase from G80/G92/GT21x to GT200/Fermi. But I'm going to wait until 7 nm GPUs before drawing any solid conclusions.
That's a lot bigger than any xx6 die from Fermi to Pascal, and similar in size to the 294 mm^2 GK104.
If this becomes a trend, then maybe Volta and Turing have ushered in a lineup-wide die size increase similar to the increase from G80/G92/GT21x to GT200/Fermi. But I'm going to wait until 7 nm GPUs before drawing any solid conclusions.