Titan V announced £2700 15TF 12nm Volta

[Wrinkly]

LOL,it seems there are some gremlins outside gaming:

https://www.theregister.co.uk/2018/03/21/nvidia_titan_v_reproducibility/

It might be worth keeping track of that in case Nvidia issues a firmware update or even replaces affected GPUs(never know owners might get a newer production batch!).

Just hope they avoid a version numbering system for the firmware

 

[N19h7m4r3]

LOL,it seems there are some gremlins outside gaming:

https://www.theregister.co.uk/2018/03/21/nvidia_titan_v_reproducibility/

It might be worth keeping track of that in case Nvidia issues a firmware update or even replaces affected GPUs(never know owners might get a newer production batch!).

Could be because despite HBM2 there's no ECC; plus it looks like the new Titan has a fork of driver seperate from GeForce, and Tesla/Quadro?

Could be the cards are too good at number crunching and are applying a bit of quantum physics into the results lol.

Could be if they overclocked them haha.

I'm so tempted to get one over my 980Ti and then overclock it. Should last me another 4+ years.

 

[Quartz]

Isn't this sort of problem well-known and caused by the conversion between binary and decimal? And being well-known, it should be well-solved.

 

[Kaapstad]

Could be if they overclocked them haha.

I'm so tempted to get one over my 980Ti and then overclock it. Should last me another 4+ years.

The cards are actually very good overclockers on both core and memory.

 

[N19h7m4r3]

The cards are actually very good overclockers on both core and memory.

The temptation is growing. Damnit. Part of me wants to try and wait for a Volta/Ampere Ti; but the other is tired of it all and wants it now.

 

[Kaapstad]

The temptation is growing. Damnit. Part of me wants to try and wait for a Volta/Ampere Ti; but the other is tired of it all and wants it now.

Wait

The Titan V is only about 35% faster than the Titan Xp in most things.

Having said that in proper DX12 it is a lot faster, sadly games don't really do genuine DX12.

 

[Rroff]

Isn't this sort of problem well-known and caused by the conversion between binary and decimal? And being well-known, it should be well-solved.

If I'm reading it right those kind of rounding error bugs shouldn't be creeping through at this level. Though they can be an issue at other times.

Maybe nVidia have accidentally implemented a true random number generator

 

[LoadsaMoney]

 

[Guest2]

Tesla V100... ~$10000

 

[james_2k]

just wait, kaap will have two in sli soon with a backup just in case one breaks!

 

[subbytna]

If we too won the lottery then we'd all have 2

 

[james_2k]

you know it!

 

[JediFragger]

just wait, kaap will have two in sli soon with a backup just in case one breaks!

He has 2, and no they can't be SLi'd.

 

[james_2k]

No he doesn’t, was referring to the v100

 

[CaptainRAVE]

Shame there are no modern games worthy of such a card.

 

[JediFragger]

No he doesn’t, was referring to the v100

My bad.

 

[james_2k]

My bad.

Probably mine too. Should have quoted the previous

 

[CAT-THE-FIFTH]

Could be the cards are too good at number crunching and are applying a bit of quantum physics into the results lol.

LOL.

Or even better, the scientific chaps may offload them if they're no use to them, so gamers can buy used ones on the cheap

One Titan V lightly used. I wonder if the ray-tracing will go slightly psychedelic though??

Just hope they avoid a version numbering system for the firmware

LOL.

Could be because despite HBM2 there's no ECC; plus it looks like the new Titan has a fork of driver seperate from GeForce, and Tesla/Quadro?

It could be that.