nvidia rtx-2080

[firejuggler]

 e cards are based on the company’s brand-new Turing architecture, initially announced last week.
The GeForce RTX 2080 Founders Edition sell for $799 and the 2080 Ti  Founders Edition will cost $1,199.
 Both are available for pre-order now  and will ship Sept. 20. The GeForce RTX 2070 will cost $599.99

https://www.tomshardware.com/news/nv...ase,37647.html


The GeForce RTX 2070 will launch in October, after the RTX 2080 and RTX 2080 Ti.

[petrw1]

I preordered the 2080Ti from NVIDIA .

[fivemack]

ECM and polynomial selection look rather unlike either huge low-precision gradient-descent calculations or ray tracing; is the RTX-2080 going to be a significant advance over the 1080 series for our mathematical porpoises?

[petrw1]

Originally Posted by fivemack

ECM and polynomial selection look rather unlike either huge low-precision gradient-descent calculations or ray tracing; is the RTX-2080 going to be a significant advance over the 1080 series for our mathematical porpoises?

Don't scare me like that ...

[tServo]

Originally Posted by fivemack

ECM and polynomial selection look rather unlike either huge low-precision gradient-descent calculations or ray tracing; is the RTX-2080 going to be a significant advance over the 1080 series for our mathematical porpoises?

Completely guessing here since there have been no performance reviews yet; I would say it's going to be faster, but not greatly faster. For a BIG improvement, a used Titan V volta would probably be the ticket. It has more memory bandwidth, probably ( but we don't know for sure ) oodles more DP FP capability, and lots more cores. Due to the cryptocurrency crash, Ebay is flooded with these boards.
Most enhancements ofthe Turing boards look geared toward gamers and pixel pushers.

[mackerel]

Assuming clocks aren't significantly different, then peak performance could scale with cuda core count, assuming software wont be using the new features directly. On that basis, I'm not expecting radical differences.

I'll leave it to those who understand the software needs and what the RTX series offers to say if there is any potential the new parts could be used to provide a useful improvement over older parts.

Quick look in ebay for the Titan Vs... one in europe, more in US, but the pricing isn't to me different enough from new to make it a consideration if it wasn't already one.

[Honza]

I've posted how RTX-2080 is doing using PG apps.

HARDWARE ⋮ RUMOR [Confirmed] NVIDIA Launching RTX 2080 Ti
http://www.primegrid.com/forum_thread.php?id=8183

Preliminary tests are complete.

MSI RTX 2080 Sea Hawk X
(GPU 1935 Mhz, temp ~50C, fan speed ~1400 RPM)

PPS Sieve: 151 - 153 s
PPS Sieve x2: 218 s for both, 109 s each
AP27: 750 sec (12m30s), twice as fast comparing to GTX 1070.

GFN15: 72-75 s
GFN16: 162-163 s
GFN17Low: 355 s; 6m
GFN17MEGA: 400 secs.; 6m40s
GFN18: 750 s (GPU load ~68%)
GFN19: 2214 s ; 37m (GPU load ~79%), twice as fast comparing to GTX 1070.
GFN20: 7210 s; 120m (GPU load ~89%), OCL4

GFN21: 6h 55m (GPU load ~94%), OCL
GTX 1060: 20 h
GTX 1070: 14 h
GTX 1080: 12 h
GTX 1080Ti: 8 h 30
TITAN V: 4 h

[M344587487]

For consumer workloads (aka gaming) the 2080 is basically a 1080ti with unused RTX functionality and the 2080ti is ~30% faster. I'd be interested to see a three way comparison for prime compute workloads. Hopefully they've worked on compute performance more than gaming performance as that's a spicy price tag to swallow for a non-validated card.

[xx005fs]

This card will be very inefficient compared to CPU because it has the useless 1/32 FP64 ratio and it will never leverage the bandwidth advantage. Titan V will definitely be currently the best value for the money for LL and PRP tests, but usually, AMD cards have really decent DP and high bandwidth so we just have to wait and see how Vega 7nm has 1/2 DP or not.

[firejuggler]

Thank you Honza for your test.

[chalsall]

Originally Posted by M344587487

For consumer workloads (aka gaming) the 2080 is basically a 1080ti with unused RTX functionality and the 2080ti is ~30% faster. I'd be interested to see a three way comparison for prime compute workloads.

Indeed. Wayne should be the one of first of us to have a 2080 in hand; I'm really eager to see how it performs for TF'ing.

As an aside, I was wondering if the 20x0's Tensor cores (basically a 4x4 FP matrix Multiply-Add in a single clock cycle) might be useful to us, but was disappointed to learn that they can't run concurrent to the CUDA cores....