• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

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

Status
Not open for further replies.
Mayh3M said:
I've asked this question a couple of times, and no one has provided an answer.

I think that's because there really is no sense to it. A powerhouse of a GPU that's going to be bottlenecked by its measly 4 GB VRAM.
Click to expand...

Because people already beating amd to release something...and because you, me, no one except amd knows if and how will it bottleneck the card...and plenty of ppl want a gpu that is strong enough to play without compromises, but not at 4K...and because they bring 8gb ones a month later.

Is that enough?
 
SiDeards73 said:
What i find hard to understand is why would AMD release their top end top tier card with 4GB of ram yet the cards below have 8GB??

Just doesnt make sense, yeah i know the HBM gen1 limitations etc, but still, why even bother?
Click to expand...

It's incredibly easy to understand.



Firstly, the overwhelming majority of games (even at 4K) are not going to be bottlenecked by 4GB, or will be bottlenecked more by lack of bandwidth than lack of capacity.

More bandwidth.

Lower latency.

Probable ability to fit larger textures into same memory.

Cards half the size - much smaller form factor and lower cost.

Much lower power than GDDR5, and much lower heat dissipation.

Way smaller memory controller = more die space for GCN units.

No more partners cheaping out with cheapo GDDR5 on non-reference.

Way simpler and vastly smaller dual GPU cards.

Clean the pipes on TSMC 28nm with the new HBM, new memory controller and revised architecture. They won't be doing EVERYTHING new on 14nmFF LP+, it'll mainly be a shrink and new foundry, with HBM2 for Arctic Islands. NVIDIA will be forced to do EVERYTHING NEW on 16nmFF, and they have a bad history with new memory controllers and memory types.

It's the future, and the sooner they get there the better. They've been developing it for 8+ years.




Why the hell wouldn't they do it? Because you're afraid NVIDIA might be in trouble because AMD's 2nd generation HBM cards will be out before NVIDIA's first? Yeah?
 
Last edited:
mururoa said:
So you are saying they will go back to using quality components for the production units, after turning their original 290X PCB into a pile of poo for the sake of cost-cutting and then bolting a 390X cooler on top if it to pass it as 390X, with no intention to use the PCB in production?

Plausible.
Click to expand...

Well maybe...AMD forbid to release information on Computex..so Powercolor maybe showing off their cooler on an old PCB to work around..
 
pmc25 said:
Probable ability to fit larger textures into same memory.
Click to expand...

Increasingly developers are already packing their textures, etc. into dds or propitiatory compression systems so little gains to be had there other than for transfer rates between parts of the GPU.

IMO they done a typical AMD and jumped on the horse too early and going to end up paying for it yet again when if they'd just waited another generation it would have been massively fruitful for them.
 
Ok so on the positive side Grenada will not just be a rebranded Hawaii, but reworked with GCN 1.2, maybe even GCN 1.3. It would have all the colour and memory compression benefits from the Tonga chip, making it quite an improvement over the currant 290x.

Seeing how well the 285 does up against the 280x with less cores, the above should put the 390x comfortably ahead of the 980, likewise the 390 ahead of the 970. They will be worth their asking prices.

################################################

On the negative side, the 390x Grenada will just be a renamed GCN 1.1 Hawaii chip, with a 50MHz core bump and a slightly bigger memory bump. It will use even bigger coolers, even some AIO units as the thermals will need to be controlled even more than the 290x.

It will be close enough to the 980 ( beating it some benches and not in others) to be a contender, A similar story for the 390 but in relation to the 970.
They will not be worth the bump in price over the currant 290/x

################################################

One of these is probably fairly close to the truth, we will know in a couple of weeks. ( well we probably wont until people actually get the cards and we can see the results of testing for ourselves, but you get the idea.):)
 
chaosophy said:
I'm still trying to work out exactly when 4GB became 'measly' ?

Though I think it was fairly recently. ;)
Click to expand...

At least it'll have the full 4gb of measly ram! People are still lapping up the gtx 970's and no ones moaning in that thread now about it being as underhanded as it was or 'measly' .

I should add, I had a 5850 which after many years of service finally had enough of my 1ghz core speed! I'm currently borrowing a gtx 680 but prior to the 5850 I had an 8800GTS so have no fanboy alliance to either side. I still am willing AMD to have pulled something amazing out of the bag!
 
Mtom said:
the moment NV used 6 and 12Gb...
Click to expand...

Doesn't really matter what NV did - you can quite easily push 4GB on high resolution gaming now - while its adequate not really what you want to see on a brand new top of the line GPU going forward - I can quite easily push 3GB on my 780 at 1440p gaming let alone anything higher so 4GB doesn't leave much headroom in the future.
 
I honestly think it just come down to R&D constraints.

Nvidia are working on HBM. We know it and AMD know it. They probably have pascal prototypes in their lab with HBM1 on board right now. The difference is nvidia can afford to spend money developing a chip that wont actually come to market whilst simultaneously developing the titan x and 980ti for this release cycle.

AMD can't afford to develop 2 completely new chips so they have to start development on HBM to avoid falling behind the curve for the next gen cards. Hence we get an expensive HBM card this year with the 4GB limit so that they are ready to do an 8GB HBM2 card when the time comes to compete with pascal.

Its the only scenario I can think of that makes sense as a GDDR5 based fiji card would be able to compete with the titan x and Ti whilst being cheaper to manufacture and not having any limits on the amount of vram available.

Thats not to say that the fury x is going to be a bad card, 4GB is still plenty for todays games at most resolutions and we know its going to be a fast card . It will be expensive though because HBM is new tech.
 
Rroff said:
Doesn't really matter what NV did - you can quite easily push 4GB on high resolution gaming now - while its adequate not really what you want to see on a brand new top of the line GPU going forward - I can quite easily push 3GB on my 780 at 1440p gaming let alone anything higher so 4GB doesn't leave much headroom in the future.
Click to expand...

Indeed, those poor gtx 970 owners with even less!
 
Jeps said:
I honestly think it just come down to R&D constraints.

Nvidia are working on HBM. We know it and AMD know it. They probably have pascal prototypes in their lab with HBM1 on board right now. The difference is nvidia can afford to spend money developing a chip that wont actually come to market whilst simultaneously developing the titan x and 980ti for this release cycle.

AMD can't afford to develop 2 completely new chips so they have to start development on HBM to avoid falling behind the curve for the next gen cards. Hence we get an expensive HBM card this year with the 4GB limit so that they are ready to do an 8GB HBM2 card when the time comes to compete with pascal.

Its the only scenario I can think of that makes sense as a GDDR5 based fiji card would be able to compete with the titan x and Ti whilst being cheaper to manufacture and not having any limits on the amount of vram available.

Thats not to say that the fury x is going to be a bad card, 4GB is still plenty for todays games at most resolutions and we know its going to be a fast card . It will be expensive though because HBM is new tech.
Click to expand...

I wish people would honestly just stop with this JUNK AMD cant afford this or that..

AM SORRY!!! But do you work for AMD?? Do you have full access to all there profit data and how much cash they have?

Am, sorry don't mean to Rage but I believe that you have ZERO! information about AMDs money and what they can or can NOT spend!
 
Jeps said:
I honestly think it just come down to R&D constraints.

Nvidia are working on HBM. We know it and AMD know it. They probably have pascal prototypes in their lab with HBM1 on board right now. The difference is nvidia can afford to spend money developing a chip that wont actually come to market whilst simultaneously developing the titan x and 980ti for this release cycle.

AMD can't afford to develop 2 completely new chips so they have to start development on HBM to avoid falling behind the curve for the next gen cards. Hence we get an expensive HBM card this year with the 4GB limit so that they are ready to do an 8GB HBM2 card when the time comes to compete with pascal.

Its the only scenario I can think of that makes sense as a GDDR5 based fiji card would be able to compete with the titan x and Ti whilst being cheaper to manufacture and not having any limits on the amount of vram available.

Thats not to say that the fury x is going to be a bad card, 4GB is still plenty for todays games at most resolutions and we know its going to be a fast card . It will be expensive though because HBM is new tech.
Click to expand...

AMD developed HBM with Hynix, they have and will be much more knowledge with it in the time pascel will be ready. NV wanted to use HMC but it fell flat on the face.
It's not AMD jumped on the bus too early..its NV who missed the bus (just as they did with the console business)

HBM is so expensive, that the cards will still be cheaper or same price than the NV counterparts even with the AIO cooler on it.
Oh and AMD already deeply in the planning of the new 14nm GPUs, as usually the GPU manufacturing process is at least 6 months before the cards release, its easy to think they close to working prototypes if not already running them...just as NV fo with the pascal.
14nm will somewhat reset the GPU race so a bit of headstart could help AMD.
 
Last edited:
Status
Not open for further replies.
Back
Top Bottom