New Titan X announced!

riotpatriot

Member
Joined
Jun 17, 2012
Messages
376
Reaction score
0
Sorry I cannot resist :(

IreCNfg.jpg
 

Gattberserk

Arch-Supremacy Member
Joined
Jan 26, 2008
Messages
10,590
Reaction score
901
If new Titan x so early announce means 490 gonna match 1080 in performance ?

Sent from Huawei NEXUS 6P using GAGT

The titan x is 2.2 times the speed of rx 480.

Is the r9 290x 2.2 times faster than 280 or 270x

Is the furyx 2.2 times faster than 380x?


All the answer will tell u if amd had any chance to stand against nvidia. If you think vega 10 will be 2.2 times faster than polaris10, i think you will need to rethink again to be realistic.
 

Gattberserk

Arch-Supremacy Member
Joined
Jan 26, 2008
Messages
10,590
Reaction score
901
The performance gap between RX 480 and Titan is just so far apart there is almost no way AMD can bridge it no matter how advance their Vega is. At least not on current GloFlo 14nm node or their current architecture.


Law of physics will tell you:

1) Double the cores will not net you double the performance. Titan X is the best example: With 50% more cores than 980, its only 33-36% faster. Law of diminishing return apply significantly here.

2) Double the cores will definitely consume double the power, or even more, since there is more current leakage due to larger information to feed etc.

3) Double the cores will almost increase the die size by 100%. This is obvious.



Based on the above 3 limits, we can safely assume that Vega need 2x the cores to compete with 1080 and 2.5x to compete with Pascal Titan X, if the architecture of Vega is identical to Polaris except the HBM controller. So assume Vega is able to pack so much...


Vega with 2x cores (1080 performance):
4608 cores, 440mm die sizes with TDP of >= 300W

With 2.5x cores (Titan P performance):
5760 cores, 550mm die size, TDP >= 375W


You can see where this is heading already... Haven talk about OC headroom and performance.


TLDR: I dont have high hope for Vega, unless something revolutionary happens with their design and architecture.
 

Gametaku

Master Member
Joined
Sep 24, 2013
Messages
4,298
Reaction score
6
The titan x is 2.2 times the speed of rx 480.

Is the r9 290x 2.2 times faster than 280 or 270x

Is the furyx 2.2 times faster than 380x?


All the answer will tell u if amd had any chance to stand against nvidia. If you think vega 10 will be 2.2 times faster than polaris10, i think you will need to rethink again to be realistic.


monikernemo is talking about GTX 1080.. nv mention Titan
 

monikernemo

Supremacy Member
Joined
Oct 23, 2014
Messages
7,268
Reaction score
1
The titan x is 2.2 times the speed of rx 480.

Is the r9 290x 2.2 times faster than 280 or 270x

Is the furyx 2.2 times faster than 380x?


All the answer will tell u if amd had any chance to stand against nvidia. If you think vega 10 will be 2.2 times faster than polaris10, i think you will need to rethink again to be realistic.
It is a mere speculation as to what triggered the release of the Pascal variant of Titan X since its barely a few months since Pascal's debut.

I don't think there's a real need to release such a card so soon since 1080 has roughly 20% improvement over Maxwell Titan X and 980 Ti as opposed to performance gains in 980 vs 780Ti.

Recall that Maxwell Titan X release was a good 4-5 months after first release of 970, 980.

Anyway, Vega 10 doesn't have to be as fast as the new Titan X to pose a threat. It just has to be faster or as fast as 1080 and sold at a much lower price. If Fury/X came out a lot earlier would Nvidia have released the first Titan X much sooner after the first Maxwell chips?
 

Gattberserk

Arch-Supremacy Member
Joined
Jan 26, 2008
Messages
10,590
Reaction score
901
monikernemo is talking about GTX 1080.. nv mention Titan

Here I have just explain a model that can compete with 1080.

Based on the above 3 limits, we can safely assume that Vega need 2x the cores to compete with 1080 and 2.5x to compete with Pascal Titan X, if the architecture of Vega is identical to Polaris except the HBM controller. So assume Vega is able to pack so much...


Vega with 2x cores (1080 performance):
4608 cores, 440mm die sizes with TDP of >= 300W


You can see where this is heading already... Haven talk about OC headroom and performance.


TLDR: I dont have high hope for Vega, unless something revolutionary happens with their design and architecture.


Please do take note the above model is taking a very very optimistic approach to cores scaling. Its not so linear in real life. 1080 having 33% more cores than 1070 and yet only perform 20-25% faster despite having better clock on top, paint a very clear picture on cores scaling.


Vega 10 will be struggling very hard just to compete with 1080. That TDP is not something I am looking forward too especially its rated stock power. That is just the 1080 performance level, and 6-9 months later.


I am sure Nvidia engineers are way smarter than me and have see this ages ago, therefore we can be safe to assume Titan X launch has nth to do with AMD. Its always been the case for Titans (OG Titan, Titan Black, TX). Nvidia releases them during the period where AMD has nothing to counter.
 
Last edited:

ragnarok95

Senior Moderator
Senior Moderator
Joined
Mar 9, 2004
Messages
125,327
Reaction score
6,588
To be honest, i really do not see Vega10 going against 1080. Like i mentioned, the most 1070. Maybe there is a 490 and a 490X. Maybe.

Vega11 will go against 1080.

If Vega11 can touch 1080 and maintain or slightly above it, it is good liao.

TitanX?? Maybe let powercolor settle it.
 

Gattberserk

Arch-Supremacy Member
Joined
Jan 26, 2008
Messages
10,590
Reaction score
901
To be honest, i really do not see Vega10 going against 1080. Like i mentioned, the most 1070. Maybe there is a 490 and a 490X. Maybe.

Vega11 will go against 1080.

If Vega11 can touch 1080 and maintain or slightly above it, it is good liao.

TitanX?? Maybe let powercolor settle it.

Actually base on the analysis I done, Titan X is renamed as Titan X is simply becz its just a beefed up version of Titan X. And why it is extremely extremely bad value for money, so much that it put previous Titan to shame.


Most of the OG TX can OCed to 1500-1550MHz given the usual condition. Now this comparison become very interesting if you compared stock pascal Titan with OCed Maxwell TX.


When you OCed OG TX to 1513MHz and compared to new TX -

1) Both have same speed
2) Both same 12GB memory (5X is just take double chunk of data but is essentially the same thing as G5)
3) Pascal TX has 16.7% more cores than OG TX.


Therefore compared both at same clock of 1513MHz, New TX is 16.7% faster, but apply the law of diminishing return with cores scaling, we know its probably just ~10% delta.


Addition note: Pascal has lower IPC per core than Maxwell (4% lower)

That put the difference between the 2 even lesser. In fact I hope someone can do a testing on OCed TX vs stock Pascal TX and see. If I am correct you should see <10% delta between the 2.


Since you can get OG TX at less than half the price of its launch MSRP at any 2nd hand market (<US$500), It might be better to wack 2 of them and OCed them hella out in SLI mode. Anyway just a suggestion, but SLI 1080 suddenly look very attractive as well. :s8:


Now the burning question remains: How much new TX can OC, given its 8+6 pins power and no AiB option.
 
Important Forum Advisory Note
This forum is moderated by volunteer moderators who will react only to members' feedback on posts. Moderators are not employees or representatives of HWZ Forums. Forum members and moderators are responsible for their own posts. Please refer to our Community Guidelines and Standards and Terms and Conditions for more information.
Top