Possible Radeon 390X / 390 and 380X Spec / Benchmark (do not hotlink images!!!!!!)

bru · 4 Jun 2015 at 14:20

So despite a lot of pages of speculation, jokes, jibes, arguments and general discussions. There is still very little that we know about this card for certain.

Definitely confirmed.

It has 4 Memory stacks. (new HBM)
It has two 8 pin power connectors (up to 375W).
It is a short card.
It has no fan.

Speculated and pretty much certain.

It will have an AIO water cooler system.
It will be faster than the 290x

Speculated and might be wrong.

Ill leave this blank as there is already 200 pages pervious to use

Troezar · 4 Jun 2015 at 14:22

humbug said:
*Server not found*

Hmm probably posting on my phone. Can someone else link it? Back to work now

Mayh3M · 4 Jun 2015 at 14:22

bru said:
So despite a lot of pages of speculation, jokes, jibes, arguments and general discussions. There is still very little that we know about this card for certain.

Definitely confirmed.

It has 4 Memory stacks. (new HBM)
It has two 8 pin power connectors (up to 375W).
It is a short card.
It has no fan.

Speculated and pretty much certain.

It will have an AIO water cooler system.
It will be faster than the 290x

Speculated and might be wrong.

Ill leave this blank as there is already 200 pages pervious to use

Its 4 GB.

SimBy · 4 Jun 2015 at 14:24

bru said:
So despite a lot of pages of speculation, jokes, jibes, arguments and general discussions. There is still very little that we know about this card for certain.

Definitely confirmed.

It has 4 Memory stacks. (new HBM)
It has two 8 pin power connectors (up to 375W).
It is a short card.
It has no fan.

Speculated and pretty much certain.

It will have an AIO water cooler system.
It will be faster than the 290x

Speculated and might be wrong.

Ill leave this blank as there is already 200 pages pervious to use

It's also the biggest die AMD ever had by far, judging from the pictures.

Squink · 4 Jun 2015 at 14:24

ashmanuk67 said:
I think they are pulling everyone together at AMD (hence lack of drivers) and they well be fooling everyone and bring out something a bit special.

Maybe they'll introduce the new cards, followed by a "just one more thing" moment, before unveiling something as pants and doomed to failure as the Shield tablet.

RavenXXX2 · 4 Jun 2015 at 14:25

There is a few ways this could go for AMD and fury.

1- It's faster than the ti/tx
2- It's on par with them
3- it's slower.

SimBy · 4 Jun 2015 at 14:26

RavenXXX2 said:
There is a few ways this could go for AMD and fury.

1- It's faster than the ti/tx
2- It's on par with them
3- it's slower.

Was this confirmed?

r7slayer · 4 Jun 2015 at 14:28

Just a thought but with Hbm would it be possible to have unified memory on one pcb? I mean four stacks of Hbm per core but unified seen as it dosnt have to use the pci-e interface to transport data. Would it be possible to have memory unified instead of mirror?

krooton · 4 Jun 2015 at 14:30

The reason people are already 'throwing AMD under the bus' is so that should the card be poor, they can shout 'I told you so!', and if it rocks, they'll keep schtum and probably buy it.

Rroff · 4 Jun 2015 at 14:31

r7slayer said:
Just a thought but with Hbm would it be possible to have unified memory on one pcb? I mean four stacks of Hbm per core but unified seen as it dosnt have to use the pci-e interface to transport data. Would it be possible to have memory unified instead of mirror?

Kind of defeats one of the bigger reasons for HBM - being close to the core for super fast transfer rates - if your running interconnects over the board it will perform fairly close to GDDR5 over the same distance.

chaosophy · 4 Jun 2015 at 14:32

One things that's bugging me is the 'won't need as much power as the 290X' comment.

Do we think that just meant the chips or the cards as a whole? And if it's true where has the power savings from HBM gone and why 2x 8pins and an AIO cooler on the top card ?

iirc the ref 290X cards where 1x8 1x6 pin?

So will it meh out the gates but flip to a 2nd BIOS and o/c massively?

Or it most of this smoke and mirrors and confusion over 390X Vs 'Fury X' ?

r7slayer · 4 Jun 2015 at 14:35

I I reckon fury card will be on par or slightly slower than a 980ti but will only feature 4gb memory. The Fury X however will have 8gb and will be noticibly faster than a 980ti or Tx
Then the dual gpu card will be the fastest single card in the world. The 390x will be slightly faster than 980 and feature 8gb
390 will be 4gb and be either slightly slower or on par with 980 but better price.

My speculation ofc

Edit - whoops fury card edit above(slightly slower)

new boy · 4 Jun 2015 at 14:36

Rroff said:
Kind of defeats one of the bigger reasons for HBM - being close to the core for super fast transfer rates - if your running interconnects over the board it will perform fairly close to GDDR5 over the same distance.

Is the problem with long interconnects that the speed of light is too slow?

Rroff · 4 Jun 2015 at 14:40

I believe the problem is the wave length rather than (just) the distance of the interconnects themselves but I'm no expert.

r7slayer · 4 Jun 2015 at 14:42

chaosophy said:
One things that's bugging me is the 'won't need as much power as the 290X' comment.

Do we think that just meant the chips or the cards as a whole? And if it's true where has the power savings from HBM gone and why 2x 8pins and an AIO cooler on the top card ?

iirc the ref 290X cards where 1x8 1x6 pin?

So will it meh out the gates but flip to a 2nd BIOS and o/c massively?

Or it most of this smoke and mirrors and confusion over 390X Vs 'Fury X' ?

The power savings have gone into powering the New core. Also wasnt 290x 2x8? Not at my pc so cnt check.

new boy · 4 Jun 2015 at 14:45

Rroff said:
I believe the problem is the wave length rather than (just) the distance of the interconnects themselves but I'm no expert.

Hmmm. Might have to have a look, sounds interesting.

Rroff · 4 Jun 2015 at 14:46

Beyond the fact that longer traces = higher latency = slower transfer rates I don't really know the mechanics involved.

Longer traces will also increase the potential for interference and parasitics you might have to deal with as well.

Geeman1979 · 4 Jun 2015 at 14:54

r7slayer said:
The power savings have gone into powering the New core. Also wasnt 290x 2x8? Not at my pc so cnt check.

6 & 8

With the exception of 2x 8 & 1x 6 on the Lightning.

Mayh3M · 4 Jun 2015 at 15:04

krooton said:
The reason people are already 'throwing AMD under the bus' is so that should the card be poor, they can shout 'I told you so!', and if it rocks, they'll keep schtum and probably buy it.

For some maybe. For others its the result of long-standing incompetence from AMD that could ultimately hurt every PC gamer in the pocket.

I have no interest in buying these new AMD cards. I have every interest in them being great performers.

ginjaninja · 4 Jun 2015 at 15:33

Nexus18 said:
Completely agree!

It makes for great reading If I had the will power and time, I would go back through some peoples post history and quote them all

Remember when the 970/980 were released and how certain people were jumping with joy because the GPU's ran quietly and cool but with the 980ti and even the TX, apparently it is a non issue now?

would be good if review sites benchmarked at various popular DB Ceilings without gpu throttling, with standards in measurement to reflect real world noise concerns.

eg up to 20db (silent for htpc use), up to 30db (in audible during game sound), upto 40db (noisy), no limit (for people with headphones or who just ant to know the limit of the product) *plucked figures out of air*..