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

[Devrij]

8gb 290 rebrand could make a really decent budget xfire setup for 1440p users

 

[gareth170]

if people really think HBM will make a huge improve will be in for a shock, it won't make much difference tbh

 

[shankly1985]

if people really think HBM will make a huge improve will be in for a shock, it won't make much difference tbh

Link?

 

[VincentHanna]

*Remembers to quote post back to gareth170 in a months time*

 

[8 Pack]

You tested this Gareth??

 

[Ayahuasca]

I think at 4k and above it will make quite a difference.

nVidia wouldn't be switching to it as well if it made no difference.

 

[humbug]

if people really think HBM will make a huge improve will be in for a shock, it won't make much difference tbh

So going through a massive GPU redesign and developing HBM to start with was a huge waste of time, should just stick with GDDR5.

Thanks for your expert input.

 

[Boomstick777]

if people really think HBM will make a huge improve will be in for a shock, it won't make much difference tbh

Yeah I mean apart from much higher bandwidth, decreased power consumption, higher overall capacity and allowing smaller design, not much difference

 

[FinglisHobbit]

I think its going to make a difference, but I almost agree with Gareth that it wont be mindblowing.

Maybe when it matures, but as the first cards to the market I'm not sure they are gonna be hugely impressive. I plan on buying one when I'm in America in July but I am not sure HBM is gonna blow DDR out of the water so quickly.

 

[whyscotty]

As above - why do it, if there are no gains/improvements to be had.

 

[humbug]

Using Hawaii, there might be a way of testing the what doubling the Memory Bandwidth could do.

I'll test the theory when i get home.

 

[Boomstick777]

It's not just that the bandwidth is massively massively increased, but that it also allows for higher capacity of memory. Lower power consumption and smaller designs.

It's a massive step forward, the extra power headroom allows for power to be used elsewhere (More shaders) while maintaining target TDP. The increased bandwidth means those extra shaders won't be held back.

This is a significant step, the biggest change in GPU architecture in sometime. The future baby

 

[gareth170]

Link?

it's what i think/feel.

uping the memory bandwidth don't give much improve. like when ocing the vram which up's the memory bandwidth don't give much performance improve.. uppng the core gives much more performance improve

 

[humbug]

If you take Hawaii as a base line Fiji is predicted to have 45% more shaders and 100% higher Memory Bandwidth.

So if one benchmarks a 290X at 1000/1250 as reference, then again at 650/625 to normalise the Core performance and Memory Bandwidth, again at 650/1250 to see what 100% higher Bandwidth actually results in scaled.

 

[gareth170]

So going through a massive GPU redesign and developing HBM to start with was a huge waste of time, should just stick with GDDR5.

Thanks for your expert input.

Yeah I mean apart from much higher bandwidth, decreased power consumption, higher overall capacity and allowing smaller design, not much difference

i meant in performance wise, which at the end of day that what we are bothered about mainly.

 

[Boomstick777]

i meant in performance wise, which at the end of day that what we are bothered about mainly.

Right but it is a big upgrade in terms of performance, it's twice as fast !

Your missing what I'm saying just having HBM brings other improvements, your getting the extra bandwidth right. Double the Bandwidth with HBM Gen 1.

2 X current bandwidth. (Big increase).

Forget that a sec, the reduced power consumption from having HBM can be spent elsewhere within the same TDP, so alongside the bandwidth just having HBM allows more shaders / transistors to be added.

So your not just getting twice the bandwidth on gen1 but other improvements as well.

Yeah it's a big upgrade, gen 1 is just the start.

 

[Dygaza]

Overclocking your memory by 10% doesn't give you even remotely good comparision of what you can actually gain by 100% more bandwith, especially if chip is designed to get good use of that.

I'm more worried that amd drivers can't keep these new cards even close to 100% gpu usage on most cpu intensive games (remember 1080p is still mainstream and a lot of ppl read those benchmarks as priority) , resulting in very mixed feeling from card previews.

Hopefully AMD can give us those improved overthead drivers they have on win10 right now. (I'm getting 20% drawcall improvements in 3dmark api test on those modded drivers at win7).

 

[Boomstick777]

It may sound strange, but PowerColor is already planning to give away two ‘yet unannounced graphics cards’.

Those cards would be Fiji’-based Radeon R9 390X models, which are expected to hit shelves next month.

The teaser only tells us that ‘The Most Advanced Graphic Card In World’ is coming soon. In the background we can see the Radeon R9 295X2 styled cooler, clearly indicating that we are expecting water-cooled graphics card.

http://videocardz.com/55539/powercolor-announces-first-fiji-giveaway

 

[FoxEye]

For those who just want the info without clicking the link.

Spoiler

Why so many low end cards? Aren't half of those less powerful than the graphics on an APU?

Also why would anybody buy a Hawaii rebrand in Q2 2015? Unless they like heat, noise and awful power consumption.

I hope this is all nonsense. AMD's rebranding merry-go-round sucks balls. They're worse than nV these days for doing it.

 

[MrMarvelous]

If the image is legit then that looks like the 295 cooler with the two gaps for tubes on the side. Pretty much confirms that the chip will run hot and AMD still can't make a decent air cooler to save their lives. Sounds like i'm ragging on it but i'm looking forward to seeing what it'll look like and perform.