GeForce RTX 2080 in TimeSpy Result Set Leaks - Titan Xp performance

When the 1070 came out, wasn't it as fast as the 980ti, and that's been the gauge for many generations, right? So this 2080 is about the same as the 1080ti, .....NOPE, no thanks.

 

As expected...

 

Story twist: It wasn't 2080 but the 2080 Ti
(Note: I'm joking, memory size and such clearly visible)

On the other hand, the card seems heavily overclocked, and -barely- manages to beat 1080 Ti. Not what I would call impressive...

 

Pretty sure chart shows a stock 1080ti not overclocked

 

Well they did go from 28nm to 16nm in the previous jump, now it's 16nm to 12nm so not an awful lot of room for optimization.

On top of that they have added all the ray tracing junk to the die, it would suggest Turing is just a 12nm Pascal plus 20% more Cuda cores & some RTX crap bolted on to it.

 

Or "cough cough" cheap.

 

Time Spy 1.0
Score 10 248
NVIDIA GeForce GTX 1080 Ti(1x) and Intel Core i7-8700K Processor
Graphics Score
10 607
CPU Score
8 601

 

Just for the reference, a similarly clocked 1080:
https://www.3dmark.com/spy/4298249 - Graphics Score 8141 points

10030 / 8141 = 23%, which is in line with what most realistic estimates from before - This series is between 15 and 25% better than Pascal. That is ALL.

Oh, and this:
https://www.3dmark.com/spy/2520142 - a similarly clocked 1080 Ti ... 10837 points. DUH!

 

Well if this is 20% faster i will go for a 2080 and sell my 1080TI that way i will get an updated warrenty a new card and more performance

 

r411.17 drivers confirmed

 

Not quite the iterative jump I'd been expecting, especially when compared to the difference between the 980 and the 1080. Disappointing.

 

So odd... i'm hoping that this was actually the 2070 or i can't see a consumer reason for this line up. Sure the 2080 is cheaper than the 1080ti, but the performance gains look to be very slim if anything.

any ETA on when reviews our out @hh or it still hush hush?

Going from this and considering its been overclocked the performance jump might be anywhere from 0-8% better, wouldn't say that is worth selling a 1080ti (considering it can also be overclocked like a champ)

 

My Ti with the same clocks as shown with 2080 have exactly the same performance in timespy test. So it's safe to say that they have the same performance. Now doing the math, a 2080 ti could be 50% faster than a 1080 Ti.

 

If anyones paid attention (not you) they'd know thats not even a mild overclock.

nor is it representative of what a fully unlocked tu104 can do.

 

Seems very disappointing to say the least unless the 2080ti has something special hidden in it... we havent seen much about it atm. but 20-25% better does not seem worth it to me, especially with the major cost increase :/ if this was a lot cheaper i could get more behind it

 

A 2080 ti have 48% more cores than a 2080. Which theoretically can reach 50% more performance than a 2080, but i wouldn't even consider this card unless playing at 4k, because at this level of power we would get cpu limited very fast at lower resolutions.

 

Didn't linus do a video of you need to basically go down to 720p/1080p to start to see any form of limit? and this is if you happen to have a 240hz monitor.... 1440p+ i assume is fine, we don't reach high enough yet

Edit:

Unless you CPU is mid-bottle end, then maybe upgrade that first

 

[H] tested a Titan V with a 7700k oced to 5GHZ and they found that at 1440p the cpu bottlenecks the card. Also that's a good read and a good prediction on how these new turing cards will perform.

 

be interesting to know if there is any 6 core/8 cores that would also bottleneck at 1440p, also how much of a bottleneck?