** NVIDIA GTX 1080 FOUNDERS EDITION: WANNA PRE-ORDER?

[Kaapstad]

The TX is not 1.5x the speed of the 980.

They are if clocked the same.

What is interesting on a coupe of tech sites is overclocked TitanXs beating overclocked 1080s.

This does tend to demolish the claim from NVidia that the 1080 is 2x the performance of the GTX 980.

 

[Five.Stars]

They are if clocked the same.

What is interesting on a coupe of tech sites is overclocked TitanXs beating overclocked 1080s.

This does tend to demolish the claim from NVidia that the 1080 is 2x the performance of the GTX 980.

I saw a review which showed a 1500mhz Titan X (yes I know they must be pretty rare) was ~10% faster than 1080 at stock speeds.

 

[drunkenmaster]

Based on transistor count compared to the 1080, the Pascal Titan is likely to have a 512 bit bus and double every thing else the 1080 comes with.

Don't think so

To do that would leave a massive amount of inactive silicon on the full fat Pascal chips.

Full fat Pascal uses over 15 billion transistors.

1.5 x 1080 would only use 10.8 billion of them.

Also with loads of cores disabled we could even see bus memory problems like those found on the GTX 970.

I'm not sure it's intentional or not but you've claimed those specs posted are wrong because the GP100 has 15.3billion transistors in it when none of those specs are based off a GP100 chip.

Then you insist because of the transistor count of the GP100, it will have double everything the GP104 has including a 512bit bus, when GP100 is already known to be HBM2, it's a design for interposer that can't work with GDDR memory at all.

This is pure nonsense, if Titan is based off GP100, it will not have a 512bit bus and it will not have double everything because we already know the exact specs of the GP100, we know shader count, we know bandwidth, we know it's HBM2.

If Nvidia launch GP100 to consumer is something that has been questioned about for the past year, if they release a higher end consumer card with no dual precision just for gaming, then it won't be GP100, and using the GP100's transistor count to decide that the rumoured specs for a potential GP102 chip is completely and utterly illogical.

 

[Kaapstad]

I saw a review which showed a 1500mhz Titan X (yes I know they must be pretty rare) was ~10% faster than 1080 at stock speeds.

That's the one.

And the problem with the FE 1080s is they are not great overclockers.

 

[Jono8]

This does tend to demolish the claim from NVidia that the 1080 is 2x the performance of the GTX 980.

Not really, Nvidia will be talking stock for stock, obviously. Whether we like it or not, Nvidia won't be comparing max overclocks to max overclocks.

Also " if they are if clocked the same" is a bit disingenuous. There is probably a higher portion of 980's that can do higher clocks than the Titan X's can.

 

[Kaapstad]

I'm not sure it's intentional or not but you've claimed those specs posted are wrong because the GP100 has 15.3billion transistors in it when none of those specs are based off a GP100 chip.

Then you insist because of the transistor count of the GP100, it will have double everything the GP104 has including a 512bit bus, when GP100 is already known to be HBM2, it's a design for interposer that can't work with GDDR memory at all.

This is pure nonsense, if Titan is based off GP100, it will not have a 512bit bus and it will not have double everything because we already know the exact specs of the GP100, we know shader count, we know bandwidth, we know it's HBM2.

If Nvidia launch GP100 to consumer is something that has been questioned about for the past year, if they release a higher end consumer card with no dual precision just for gaming, then it won't be GP100, and using the GP100's transistor count to decide that the rumoured specs for a potential GP102 chip is completely and utterly illogical.

Please enlighten us with what you think the spec will be then.

 

[Kaapstad]

Not really, Nvidia will be talking stock for stock, obviously. Whether we like it or not, Nvidia won't be comparing max overclocks to max overclocks.

Using max overclocks is the only fair way to do it.

What is stock for Maxwell ?

A reference card, an AIB partner card, a Kingpin even. There are so many variables that using max overclocks removes.

 

[Jono8]

Using max overclocks is the only fair way to do it.

What is stock for Maxwell ?

What Nvidia's specification is obviously

 

[Boomstick777]

There really needs to be some positive news on the 1070 soon to counter all the 1080 negativity. Has definitely pi**ed a lot of people off, but if the 1070 necks the 980Ti at a fair price, we'll have some peace (hopefully). I am still expecting a £400+ price for that though, even the AIB custom cards, but would love to be wrong.

Agreed, really hope it comes in at a fair price. I wanted to return to the green team until Vega at the end of the year. 1070 would be spot on until then. If it's priced as badly as the 1080 then I'll keep using the Nano. That's not a bad thing, this card has lasted me the longest out of all of em so far . If 1070 isn't a rip off will pick one up at launch and gonna pass the Nano onto my boys.

 

[Exibit]

It'll all depends on who gets some initial stock. It's no good a company having the lowest price but no stock for weeks after launch. At least OC have said they will have some stock on day 1, will the others?

Despite the pricing I think many will be disappointed if the initial stock levels are going to be low - they'll probably sell out in a short time.

I have checked that company is also having units available the day they available to OC

 

[D.P.]

Don't think so

To do that would leave a massive amount of inactive silicon on the full fat Pascal chips.

Full fat Pascal uses over 15 billion transistors.

1.5 x 1080 would only use 10.8 billion of them.

Also with loads of cores disabled we could even see bus memory problems like those found on the GTX 970.

Full fat Pascal, as in GP100 has a lot of transistor for FP64 compute that will do nothing for gaming. The GP100 is DSR from twice the gaming GPU of the GP104. A GP102 that is a scaled up GP104 will have far better gaming performance than GP100 in a smaller, cheaper, cooler die.

 

[D.P.]

The TX is not 1.5x the speed of the 980.

Nope, but it had 1.5x the shader capacity and bandwidth.

 

[NirK]

I am surprised that you think the nvidia 1080ti will have 5000+ cores, with 20,000 posts you obviously read a lot about graphics cards etc.... have you not noticed a pattern with graphics cards? The 1080ti will be around 3500 cores, 384 bit etc. the titan around 3800 and more memory.

Yes they physically could make a 600nm chip with 5000+ cores etc. but they won't... 1080ti / titan will be about 30-50% faster than the 1080... I would be extremely surprised if they have 5000+ cores I think it is extremely unlikely that they will release a card with 5000+ cores!

 

[D.P.]

I am surprised that you think the nvidia 1080ti will have 5000+ cores, with 20,000 posts you obviously read a lot about graphics cards etc.... have you not noticed a pattern with graphics cards? The 1080ti will be around 3500 cores, 384 bit etc. the titan around 3800 and more memory.

Yes they physically could make a 600nm chip with 5000+ cores etc. but they won't... 1080ti / titan will be about 30-50% faster than the 1080... I would be extremely surprised if they have 5000+ cores I think it is extremely unlikely that they will release a card with 5000+ cores!

Agreed.

 

[NirK]

Would be great though if Nvidia would just make a 5000 core monster card!!

 

[[HB]Rugrat]

Would be great though if Nvidia would just make a 5000 core monster card!!

Yeah but given their pricing strategy it'd cost about eight grand.

 

[D.P.]

Would be great though if Nvidia would just make a 5000 core monster card!!

The thing is game performance simply won't scale that well to the extra cores. If you look at Nvidia's GPU evolution they certainly don't double the number of cores each generation, in fact there are big performance gains when the number of cores goes down actually. Keeping the cores properly utilized and r,roving all bottle is in the pipeline is the most important t thing to performance.

 

[NirK]

The thing is game performance simply won't scale that well to the extra cores. If you look at Nvidia's GPU evolution they certainly don't double the number of cores each generation, in fact there are big performance gains when the number of cores goes down actually. Keeping the cores properly utilized and r,roving all bottle is in the pipeline is the most important t thing to performance.

I think they stick to something like xx70 at last years ti, xx80 bit faster than the old ti, xx80ti and titan about 30-50% faster...repeat

1080ti should be very fast, I think those leaked specs are probably right... but then we will probably get 120hz 4k monitors on DP 1.4 and people who want 4k 60fps now, will then want 4k at 100 FPS with HDR or 8k or something else lol.

 

[Boomstick777]

I am surprised that you think the nvidia 1080ti will have 5000+ cores, with 20,000 posts you obviously read a lot about graphics cards etc.... have you not noticed a pattern with graphics cards? The 1080ti will be around 3500 cores, 384 bit etc. the titan around 3800 and more memory.

Yes they physically could make a 600nm chip with 5000+ cores etc. but they won't... 1080ti / titan will be about 30-50% faster than the 1080... I would be extremely surprised if they have 5000+ cores I think it is extremely unlikely that they will release a card with 5000+ cores!

Yeah expecting the Ti to have around 3500 CUDA CORES, and the Titan maybe a few more. If they stick with GDDR5X on the Ti / Titan brand I will be shocked, such cheap skates lol. They will still price them to the moon as well.

 

[robbw77]

Any confirmation on the pre-order?

If we can on Friday, will it get here for Saturday?