TPU: NVIDIA to Unveil GeForce GTX TITAN P at Gamescom

Page 3 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

LTC8K6

Lifer
Mar 10, 2004
28,520
1,575
126
TechPowerUp highlights just how severe the CPU bottleneck is now becoming at "lower" resolutions such as 1440p. Even Skylake i7 6700K 4.5Ghz cannot keep up with 1080 SLI.

...

This is the difference between buying PC parts for bragging rights (I spent more $$$ on hardware so I am a "PC enthusiast") and being a true PC enthusiast.

A true PC enthusiast who prioritizes gaming performance will admit that i7 6700K @ 4.8Ghz and DDR4 4000-4133 will smash a 4.4Ghz i7 6950X with DDR4 3000 and get the cheaper but faster platform. And of course there are already Z170 boards (1, 2) that can support DDR4 4500mhz.

It's good to see NV actually call out the truth and recommend PC gamers buy the fastest gaming CPU to extract maximum performance out of the 1080 SLI/Titan P (etc.).

By the time we are into Volta generation, CPU bottlenecking will be even more severe. This is why both AMD/NV are pushing 1440p 144-200Hz, HDR, 4K/5K gaming. 1080p 60Hz gaming is becoming a completely outdated concept for the level of graphics cards that are coming out now and unless games start becoming a lot more advanced, 1080p 60hz gaming will be laughable by 2018 when a $400 card will be as fast as a GTX1080Ti. GPU hardware is now starting to outpace software as we enter the 2nd half of the current console generation. We've seen the exact same scenario play out towards the last 3 years of PS360 generation.

Where does this leave 8C/16T Zen?
 
Mar 10, 2006
11,715
2,012
126
This has been gone over more that a few times, but so would I. The only way it would make sense to launch a GP100 based Titan before the PCIe based Tesla cards is if they have a large number of dies that can't even make P100 spec, and its a further cut part. 12GB, 3328CC (or 3200), and 1/4 or 1/3 DP performance.

Still, I can't see them launching that before the Tesla cards.

I just think a Tesla P100 PCIe, even in its highest end config, would only be marginally better than a GTX 1080 for games.
 

railven

Diamond Member
Mar 25, 2010
6,604
561
126
Welps, I get the feeling if AMD doesn't respond, as another poster here said we won't see anything based on big Pascal outside of Titan for a while.

Big Kepler didn't see the world as a consumer product until almost a year later and it was the flagship for a new generation. And Big Maxwell only came out as a gaming product only when Fury actually seemed to bring some heat.

If the power ratings for that 1480mhz OC'd RX 480 is any indication of what AMD has in store for us, woof.

Am I going to have to pull a Sony/PS3 and get another job to afford my next GPU?
 

sub-80

Senior member
Jan 11, 2014
250
0
76
Titan "P" does not sound exciting hopefully the performance is above a slight enhancement from the Titan x

Sent from my SM-G930F using Tapatalk
 

ultima_trev

Member
Nov 4, 2015
148
66
66
Hasn't there always been at least a six month delta between the mid sized x80 and the full size Titan / x80 Ti (i.e. Kepler and Maxwell)? It seems too soon for the big Pascal.

I'm not saying it isn't possible, but I figure nVidia would want to milk consumers with 1080 / 1070 as much as possible before cramming whatever Titan / 1080 Ti SKU down people's throats.

Also, did nVidia officially announce there would be a GP102 chip to slot between GP104 and GP100? I see a lot of speculation and rumors, but where is the source?
 
Last edited:

alcoholbob

Diamond Member
May 24, 2005
6,271
323
126
Hasn't there always been at least a six month delta between the mid sized x80 and the full size Titan / x80 Ti (i.e. Kepler and Maxwell)? It seems too soon for the big Pascal.

I'm not saying it isn't possible, but I figure nVidia would want to milk consumers with 1080 / 1070 as much as possible before cramming whatever Titan / 1080 Ti SKU down people's throats.

Also, did nVidia officially announce there would be a GP102 chip to slot between GP104 and GP100? I see a lot of speculation and rumors, but where is the source?

It could also be that Nvidia believes that AMD might launch Vega this year and wanted to get the Titan out first. And then after they get an idea of where Vega will be they will tweak the 1080 Ti/cut down card.
 

Grooveriding

Diamond Member
Dec 25, 2008
9,107
1,260
126
Only a month until gamescom. Availability is so poor on the 1080 that the only thing I ever see in stock is the reference card with the extra $100 premium for the privilege of buying the worst model of the 1080. EVGA still hasn't even shown the Classified or listed it on their site. I usually just get reference for their mid range cards and a water block, but have been waiting on the Classified because it comes with water block support. At this rate though I'll be able to just get a Titan P and not even bother with nvidia's mid range card this cycle :p
 

Sweepr

Diamond Member
May 12, 2006
5,148
1,142
131
Over at Beyond3D two users (including one that also posts here) claim to have heard GP102 is 384-bit GDDR5X, ready for back-to-school season (Q3). Expected to reach frequencies very close to GP104, 35-45% faster in GPU bound situations. ;)
 

DooKey

Golden Member
Nov 9, 2005
1,811
458
136
Over at Beyond3D two users (including one that also posts here) claim to have heard GP102 is 384-bit GDDR5X, ready for back-to-school season (Q3). Expected to reach frequencies very close to GP104, 35-45% faster in GPU bound situations. ;)

Sign me up.
 
Mar 10, 2006
11,715
2,012
126
Over at Beyond3D two users (including one that also posts here) claim to have heard GP102 is 384-bit GDDR5X, ready for back-to-school season (Q3). Expected to reach frequencies very close to GP104, 35-45% faster in GPU bound situations. ;)

Sexy. This is very good execution from NVIDIA, they aren't giving their competition any room to breathe.
 

24601

Golden Member
Jun 10, 2007
1,683
39
86
Sexy. This is very good execution from NVIDIA, they aren't giving their competition any room to breathe.

It's annoying that if those numbers are true, they are leaving massive performance on the table with huge memory bandwidth bottlenecks for a refresh probably a year after release for each card.

(As is currently the case for all Pascal cards released so far other than perhaps 1060)
 
Mar 10, 2006
11,715
2,012
126
It's annoying that if those numbers are true, they are leaving massive performance on the table with huge memory bandwidth bottlenecks for a refresh probably a year after release for each card.

As Micron can get higher speed GDDR5X into production, I can totally see a Pascal Refresh in 2017 if Volta isn't ready by then.
 

linkgoron

Platinum Member
Mar 9, 2005
2,286
810
136
Sexy. This is very good execution from NVIDIA, they aren't giving their competition any room to breathe.

nvidia doesn't need to release GP102 for that. They just need to release the 1060 6gb at 250$. At the high end AMD haven't even matched AIB 980tis, and we're more than a year after the release of the 980ti. The "best" rumors say that it will take at least 4 more months (or six+, if Vega is 2017, which I believe it is) to match it, and _maybe_ also match the 1080.

I don't really understand why nvidia needs to release GP102 (/titan) so soon. The 1080 is still very scarce, stock is still not that easy to come by, demand is high and AMD have nothing which is even close to competition.
 

24601

Golden Member
Jun 10, 2007
1,683
39
86
nvidia doesn't need to release GP102 for that. They just need to release the 1060 6gb at 250$. At the high end AMD haven't even matched AIB 980tis, and we're more than a year after the release of the 980ti. The "best" rumors say that it will take at least 4 more months (or six+, if Vega is 2017, which I believe it is) to match it, and _maybe_ also match the 1080.

I don't really understand why nvidia needs to release GP102 (/titan) so soon. The 1080 is still very scarce, stock is still not that easy to come by, demand is high and AMD have nothing which is even close to competition.

Probably a JHH pride thing. Nvidia does alot of things that investors hate.

Things like the magnesium alloy Titan cooler theme are purely for pride, they could have easily just made it aluminum from the start and no one would have noticed other than Nvidia.
 
Last edited:

tviceman

Diamond Member
Mar 25, 2008
6,734
514
126
www.facebook.com
It's annoying that if those numbers are true, they are leaving massive performance on the table with huge memory bandwidth bottlenecks for a refresh probably a year after release for each card.

(As is currently the case for all Pascal cards released so far other than perhaps 1060)

I said this before GTX 1080 launched. Despite all the noise Nvidia has made about color compression to save bandwidth, GTX 1080 and GTX 1070 are still bandwidth starved, which is part of the reason why GP106 was designed with 1/2 the shader power but 75% of the bus and ROP functionality of GP104. Expect 1100 series refreshes next year with the same relative core clocks and faster vram (12 ghz for GP102 and GP104).