Originally Posted by DarkIdeals
Even your own cherry picked review doesn't agree with those numbers...sigh.Warning: Spoiler! (Click to show)
Dragon Age Inquisition 39% increase with TITAN X over 980
Shadow of Mordor ~37% increase with TITAN X over 980
GRID Auto 36% increase with TITAN X over 980
Far Cry 4 35% increase with TITAN X over 980
And in other reviews you'll see it floating much more towards 40% and often close to 43-45% in some games. On average the TITAN X maxwell is close to 40% over the 980. You're trying to claim that this new TITAN won't be nearly as big of a difference simply due to a slight change in cuda-core percentage which is ludicrous. That "10%" is only 256 cuda cores, which we should ALL be well aware of the fact that "when on the same die" a difference of 256 is exactly the difference between the Maxwell TITAN X and 980 TI! And EVERYONE and their mom was saying "TITAN X isn't worth anything, 980 TI made it UTTERLY redundant" because there was literally like 3-4% difference AT MOST between the two. And using your same argument of "percentages" i can point out that 3,840 (50% over 1080 core-wise) vs 3,584 (40%) is even SMALLER of a real world difference than it was in the 980 TI vs Maxwell TITAN X comparison due to the larger starting numbers. 3072 vs 2816 (TITAN X vs 980 TI) was only a ~9% difference if you divide larger by smaller; do the same for 3840 vs 3584 and you get 7%. This means that the ~3-4% real world difference between TITAN X and 980 TI will actually be LARGER than the difference between THIS card and the theoretical "full GP102 3840 core" meaning only ~2% difference!
This means you are going to see AT MOST a ~2-3% lower difference between 1080 and this TITAN than we saw from 980 to Maxwell TITAN X! That is just pure math.
The fact of the matter is, this TITAN X is STILL 1024 cores more than the x80 model; which is the EXACT same as the 1024 cores that the 3072 core Maxwell TITAN X had over the 2048 core 980. That is why it will perform very close to identical to how the TITAN X did over the 980 in this comparison of 1080 vs TITAN XP.
Cherry picked? Oh right because 39%, 36.6% is sooooooooooo different from 35%.
Seriously who's the one doing the cherry picking now? Btw go back and check your math with GRID, it's only 34%, not 37% like you're trying to claim.
What "other reviews" are you referring to? Because TechPowerUp found even at 4K, the average difference between a Titan X and 980 is 31.6%.Warning: Spoiler! (Click to show)
And in Guru3D's review
, the absolute best it does @4K is 39.3% over 980 in Hitman Absolution, and that's just one game. The average is still around 35%. HardOCP
didn't test 4K, but the absolute best they got was still 35.6% over 980 in Dying LIght a 1440p.
If I really wanted to cherry pick, I could've picked Hexus's
review, because literally every single game they tested the Titan X fell short of 35% over 980 at 4K -- the best they got was 34% in Tomb Raider, and the worst they got was only 23.8% in Shadow of Mordor. If anything I "cherry picked" one of the most favorable reviews, and I'm supposed to be the one doing the cherry picking.
And you COMPLETELY missed my point. Titan X had 50% more of everything than 980, and while yes there was on average a 6-7% deficit in boost speed compared to 980, the average improvement still ended up being around 35%, even at 4K
. Even if we add back that 7% to 35% (and that's assuming perfect 1:1 scaling), with 50% more of everything, clock for clock Titan X would still be 42% faster than 980 on average, using unrealistic math.
Titan XP only has 40% more shaders than 980, and the listed boost clock of Titan XP is only 1513 MHz, which is already 12.7% lower than 1080's listed boost of 1733 MHz. And you somehow expect Titan XP to be 35% faster than 1080, stock vs stock.
Originally Posted by Klocek001
now it has just come to me, why is he comparing a full chip monster like 980Ti with a 1080. They're not the same tier. High-end vs "mid-range +" class. It's like measuring rx480 vs 390X for IPC, 390X is gonna win clock for clock, with a huge lead.
I think TXM vs 980 is gonna be better than TXP vs 1080. TXP is cut down, TXM is a full die. Additionally, I think 1080 is pretty fast for x80 card. For example, let's see 380X CF vs 980 and RX480 CF vs 1080.
Warning: Spoiler! (Click to show)
It took a 980Ti for Maxwell generation to beat two R9 380X cards, now 1080 is beating CF RX480.
Originally Posted by guttheslayer
The performance gap between TXP and GTX 1080 is the same as GTX 980 to 980 Ti.
Therefore the gap between TXM vs 980 is definitely bigger than 1080 to TXP.
Yeah well tell it to the guy I'm responding to.Edited by magnek - 7/28/16 at 9:37am