Pascal Titan-X Launch

[Gerard]

Reviews should be interesting, clocking the core frequency down and relying on the extra cuda cores to compensate smacks of the same thing they did with titan z.

 

[straxusii]

Unless they clock really well it's going to be carnage in the reviews. Core count, frequency and tflops show it only 25% faster than a 1080

 

[Gerard]

Unless they clock really well it's going to be carnage in the reviews. Core count, frequency and tflops show it only 25% faster than a 1080

Yet around (we think) double the price, total bargain!

 

[RavenXXX2]

Unless they clock really well it's going to be carnage in the reviews. Core count, frequency and tflops show it only 25% faster than a 1080

It's not for gaming.

 

[Gerard]

It's not for gaming.

Sure its not. Like the last few titan's weren't for gaming either. As long as it has the "GeForce" branding, its a gaming card, end of.

 

[straxusii]

It's not for gaming.

Anyone know what it's FP64 performance is, if it's crippled like the maxwell Titan X it won't be for compute either

 

[modd1uk]

Interested to see the real world performance, will it justify the cost? Like the other titans I feel it will be a " LOOK AT MY BAWLS " card, the elite card owners club.

 

[D.P.]

Unless they clock really well it's going to be carnage in the reviews. Core count, frequency and tflops show it only 25% faster than a 1080

If you look at all the numbers the Pascal Titan has a larger theoretical gap compared to the 1080 than the Maxwell TX had to the 980. No one complained about that.

 

[Dirk Diggler]

Are we expecting reviews/benchmarks today?

 

[straxusii]

If you look at all the numbers the Pascal Titan has a larger theoretical gap compared to the 1080 than the Maxwell TX had to the 980. No one complained about that.

Not sure what you mean?

Pascal TX to 1080 is 11tf vs 8.8, ie 25% faster.

Maxwell TX to 980 is 6.144tf vs 4.61, ie 33% faster.

 

[D.P.]

There is much more to a GPU than the theoretical terraflops. have a look at Pixel fillrate, texture fillrate, memory bandwidth, geometry engines, cache etc.

 

[straxusii]

There is much more to a GPU than the theoretical terraflops. have a look at Pixel fillrate, texture fillrate, memory bandwidth, geometry engines, cache etc.

Of course, the number of cores and the clock speed compared to 1080 doesn't fill me with confidence either. Seems to be about +25% also.

 

[Greebo]

There is much more to a GPU than the theoretical terraflops. have a look at Pixel fillrate, texture fillrate, memory bandwidth, geometry engines, cache etc.

Agreed and certain bigger increase might make the Titan perform better in certain circumstances but overall, esp from the same generation, terraflops is a perfect way of judging likely performance.

 

[Telecaster]

if it clocks like the 1080 under water then I will find it hard to resist, however if it turns out just to be a failed chip with the clocks knocked back then I think I'll be waiting.

 

[The-Hoff]

The only info you can find on this card is the 'Synthetic Benchmark Results' posted on every tech site out there and I am fed up of seeing it.
I don't believe it will be out tomorrow for one minute.

 

[NirK]

The only info you can find on this card is the 'Synthetic Benchmark Results' posted on every tech site out there and I am fed up of seeing it.
I don't believe it will be out tomorrow for one minute.

But the reviews etc. will be ready, then they have to wait until the agreed date (NDA) until they can put them online. The reason there is nothing online is because they are not allowed.

With the actual Titan X vs 1080, I think the Titan X could potentially be very good or a flop, this depends on the boost clocks, if these cards can boost to 2Ghz like the 1080 then it will be very good, if it can only go to 1600mhz or something, then it will be a flop, but some people will watercool them etc. and probably they will be good then, but I have a feeling that on the reference cooler it will be a bit of a flop. Obviously it will be faster than 1080, but not really worth twice the MSRP. 25% faster than 1080 for double the MSRP, but 40% faster would be good.

 

[Besty]

reviewers have cards, expect nda to lift tomorrow Pacific time with some dodgy Foreign site to post some crappy inconclusive benchmarks overnight using programs nobody cares about

 

[Ian Evey]

I do wonder if there will be a Titan-FE-X card first.

 

[Besty]

nah just the Titan x scve (slightly cripped until vega edition)

 

[D.P.]

Agreed and certain bigger increase might make the Titan perform better in certain circumstances but overall, esp from the same generation, terraflops is a perfect way of judging likely performance.

No, terraflops is a measure of theoretical fp32 computing resources. Considering the GP102 already distorts that image with int8 and FP16 support you already have a poor baseline. Then there is the fact that such a card is aiming for the 4K crowd where things like texture filtrate and bandwidth become more critical. Having 50% more bandwidth is more important than the compute aspects.