Chiphell Fiji xt/Bermuda xt/gm200

For those in need of translation, they are speculating that:

1. Those are made from 20nm process by GlobalFoundries
2. The flagship card is not Fiji XT, but Bermunda
3. Bermunda is faster than 290x by 65%, and ref uses Hybrid cooling (water and air).
4. GM200 engineering sample beats 980 by 34%


Proceed with caution and a sea of salt.
 
Salt taken, but if those numbers are true AND that's a single GPU card... Daayuum!

GM200 and Bermuda should be pretty close in performance at stock.

I do have a nagging feeling that AMD will be going the "ultra-premium" route with Bermuda and pricing in the $1000 range :(
 
If that really is the case, then AMD has a winner, considering the 390X is even faster than the full 980 Ti.

but yea, salt and all.
 
I love salt, especially the salty kind. Just want to know price- hoping fro 550 so i cant get the step down 390.
 
Considering current 20nm yields, the liquid cooling and the performance and power/heat advantage this will have (if rumors are true), the price will likely be commensurate. I'm guessing $700-1000.
 
ugh, i dont want no stinkin CLC.

Agreed, I wouldn't buy a card if that is the only option for cooling. So hopefully there will be some nice air cooled versions from the usual suspects.

Sounds like a beast of a card though. Let's hope the water isn't too salty.
 
Would be great if AMD would also make a deal to supply full WC blocks on their premiums cards along side the CLC coolers.

Tired of dealing with air coolers, liking the water, a lot.
 
Just like that I canceled order of gtx 980 for $349 with dem cashbacks. Chiphell better be true lmao.
 
anything about ram? if this is anything close to the truth it makes 4k gaming at higher graphics settings a reality. if they ship these cards with 4gb i am going to bang my head against a wall wondering "WHY???"
 
I hope these are real!...and very interesting turn of events if AMD can truly take the single GPU performance crown back; maybe it will keep Nvidia honest in it's pricing of the GM200. The large GM200 die could be a beast underwater too assuming they had to underclock it compared with the 980 due to the extra cores.
 
Why are they putting the 390X on water? It can't possibly be necessary.
Unless this is AMD's way of trying to attract flagship buyers... You get a super fast card, with water, for a good price (hopefully). An added bonus.
 
Why are they putting the 390X on water? It can't possibly be necessary.
Unless this is AMD's way of trying to attract flagship buyers... You get a super fast card, with water, for a good price (hopefully). An added bonus.

To combat the negativity that previous ~+250w cards received due to temps and acoustics.
295x2 received some very good reviews due to their hybrid solution.
 
Why are they putting the 390X on water? It can't possibly be necessary.
Unless this is AMD's way of trying to attract flagship buyers... You get a super fast card, with water, for a good price (hopefully). An added bonus.

They couldn't make a dual chip without water, they might have trouble cooling a single chip now. If they are just making another Hawaii chip with more transistors it's probably molten hot.
 
They couldn't make a dual chip without water, they might have trouble cooling a single chip now. If they are just making another Hawaii chip with more transistors it's probably molten hot.

what if they are doing to allow for higher overclocking? just because it isnt made by nvidia, doesnt mean it isnt good.
 
They couldn't make a dual chip without water, they might have trouble cooling a single chip now. If they are just making another Hawaii chip with more transistors it's probably molten hot.
And this is why it's a bad idea to ship a card with factory watercooling.
People take one look at it and assume it runs hot. Especially after the 295X2.

I don't think AMD's marketing team is very clever.

There are people running GTX 970's and 980's on water, and those aren't "molten hot".
And then... there's this.
 
People forget, the ~50W TDP decrease on the 300 series cards is coming from HBM RAM. The chip itself is probably still a furnace, which is why watercooling makes sense.
 
And this is why it's a bad idea to ship a card with factory watercooling.
People take one look at it and assume it runs hot. Especially after the 295X2.

I don't think AMD's marketing team is very clever.

There are people running GTX 970's and 980's on water, and those aren't "molten hot".
And then... there's this.

I don't think they could have made the 295x2 without watercooling. It's probable they won't be able to make a chip to compete with Maxwell without watercooling. The 290x is pretty much at the limit of air cooling and that's with throttling that messes with performance.
 
I don't think they could have made the 295x2 without watercooling. It's probable they won't be able to make a chip to compete with Maxwell without watercooling. The 290x is pretty much at the limit of air cooling and that's with throttling that messes with performance.

The leaked 380x benchmarks are faster then the 980 and that isn't watercooled.
 
This thread is like taking a time machine back to 2013.
Why are we still talking about Hawaii?

http://www.chiphell.com/thread-1182382-1-1.html

The "380X" there is faster than the 980 and better per watt.
Even if that's the 390X, I doubt they need water for a whopping 197 watt card.

If they do have a 250W card on water for the 390X then it's safe to assume it's going to annihilate the GTX 980 since the 380X is already doing so.
Time to throw those 290's in the garbage, I know that's where my tahiti is going.
 
I am a bit worried that the release of anything is far from now. Usually we get some decent rumors to go off of 2-3 months before something comes out, but it has been pretty quiet.
 
how the hell does one person get 4 engineering samples..... from both camps?


It's called psychotic episodes. The mysterious powers of the mind unleash upon oneself a slew of fantasy scenarios that are then believed into creation and produced as truth instead of imaginary.
 
The 380X is Hawaii 2.0 reboot using a different fab but still 28nm and power savings vs TMSC 28nm .. if this is the road AMD plans for here on out then Nvidia will not have anything come to easy anymore if we only look at 290x vs 380x as that is still Hawaii on 28nm.. 380x may have a lot of overclocking headroom because of much power Hawaii can eat on TMSC 28nm.

I called 50% performance for 290X > 390X .. hate that I am wrong..
 
Last edited:
One interesting effect if AMD moves away form using TSMC for GPU releases could mean a even more staggered release schedules from both camps (more so than the last few years even). By this I mean people may need to reconsider what constitutes the same generation.

Let's say in a few months Fiji/Bermuda ends up leading for 6-9 months, but then Pascal (GPxxx?) ends up leading Bermuda for 6-9months, then AMD releases something to lead 6-9 months, then back to Nvidia. This could be the type of scenario going forward.

how the hell does one person get 4 engineering samples..... from both camps?

The interesting thing about AMD and Nvidia is for GPUs they largely share the same partners (from supply to manufacturing). 4 of the 5 largest video card companies actually sell from both camps for example.

It's almost certain that AMD and Nvidia know much more in advance what each other is likely to do than the internet rumor mill.
 
How do you know it's not watercooled? Or real?

Normal people want those bences to be real because of competition, low prices and so on (in theory), while fanboys of either camp want their favorites to pawn the other ones. Simple psychology :D .
 
Back
Top