**AMD Radeon 390X Graphics Card WCE**

[Silent_Scone]

There is a lot of misunderstanding about the actual limitation of HBM, and it seems that there is no difference here either. People have misunderstood Drunkenmaster on this one


Layman's:

Is it possible to fit more than 4gb on a single interposer? Potentially

Will have AMD have done it? Potentially

Is it possible to have scrapped 4gb and gone with 8gb in recent months given events? Near on impossible

Is HBM limited to 4gb? No it is limited to four dies per STACK there is no reason within the spec why adding additional stacks isn't possible, it just depends on die area and other afflicting limitations

Is the fud article nonsense then? 100% yes

 

[SiDeards73]

I said this ages ago, 3 tiers of dx12 and R9 290 etc are tier3 while no current Nvidia card is, probably due to dx12 being designed for the xbox1 which has AMD hardware.

This is why holding on to a 290 right now is not a bad idea til the die shrink and HBM 2nd gen cards come, as once dx12 starts seeing daylight in games older supporting hardware will get a massive lease of life

 

[Mei]

whats with having tier lists for everything?
im bored of tier lists in HOTS alrdy! i dont want microsoft to start up on it ><

 

[Final8y]

There is a lot of misunderstanding about the actual limitation of HBM, and it seems that there is no difference here either. People have misunderstood Drunkenmaster on this one


Layman's:

Is it possible to fit more than 4gb on a single interposer? Potentially

Will have AMD have done it? Potentially

Is it possible to have scrapped 4gb and gone with 8gb in recent months given events? Near on impossible

Is HBM limited to 4gb? No it is limited to four dies per STACK there is no reason within the spec why adding additional stacks isn't possible, it just depends on die area and other afflicting limitations

Is the fud article nonsense then? 100% yes

+1

If there is an 8GB version then thats because it was already in the works and not a last minute change.

 

[Kaapstad]

There is a lot of misunderstanding about the actual limitation of HBM, and it seems that there is no difference here either. People have misunderstood Drunkenmaster on this one


Layman's:

Is it possible to fit more than 4gb on a single interposer? Potentially

Will have AMD have done it? Potentially

Is it possible to have scrapped 4gb and gone with 8gb in recent months given events? Near on impossible

Is HBM limited to 4gb? No it is limited to four dies per STACK there is no reason within the spec why adding additional stacks isn't possible, it just depends on die area and other afflicting limitations

Is the fud article nonsense then? 100% yes

Shame DM did not understand this in some of his older posts.

 

[flopper]

390x fast and cool
Godlike and I am worthy.

 

[DavidS14]

I think that AMD looked at both the 4GB and 8GB and did the development for both with an eye to bring the 4GB card out first and then later add the 8GB card to the line up as we've seen before.

With both cards probably ready they have looked at the market and decided that with the push for more VRAM they should bring the 8GB out first and maybe add a cheaper version later. There's no way given the time frame that they would swap to 8GB if all of the development was not already finished or near to being finished. It would be short sighted not to plan for both VRAM capacities and if you have then they would be concurrent as you would want as much cross over as possible to save money.

 

[Kaapstad]

I think that AMD looked at both the 4GB and 8GB and did the development for both with an eye to bring the 4GB card out first and then later add the 8GB card to the line up as we've seen before.

With both cards probably ready they have looked at the market and decided that with the push for more VRAM they should bring the 8GB out first and maybe add a cheaper version later. There's no way given the time frame that they would swap to 8GB if all of the development was not already finished or near to being finished. It would be short sighted not to plan for both VRAM capacities and if you have then they would be concurrent as you would want as much cross over as possible to save money.

This 100%

 

[eddyr]

Well daym!! PDF, it must be legit right??

If there was a change to the launch schedule to accommodate an 8GB model then it was always on the horizon and it's simply a good decision to launch both 4/8GB simultaneously. I eagerly await to read reviews about new stuff.



--------------------------
As an aside fellas, let's not be disingenuous about what has been said and when it's also remarkably quick and easy to check.


http://forums.overclockers.co.uk/showthread.php?p=27747332#post27747332
http://forums.overclockers.co.uk/showthread.php?p=27747987#post27747987

 

[SciNeT]

Functionality spelled functionallity on that slide, is this a spelling error AMD would make?

 

[TheRealDeal]

390x fast and cool
Godlike and I am worthy.

Who else is worthy?

 

[flopper]

Who else is worthy?

Many are called
A few hears the call
The mirror on the wall tells it all

 

[TheRealDeal]

Yup, Nvidia working on DX12 and closely with Nvidia for 5 years didn't they try and say... yet GCN is ostensibly a DX12/Vulkan/Mantle architecture and yet Maxwell which came out after DX12 had been announced has not huge dx12 support while previous hardware has even less DX12 support. Nvidia apparently working so closely they forget to actually add it to their hardware. In fairness, they've done that before DX10, missed out some features then begged MS to remove them from the DX10 spec then paid for DX10.1 updates which speed up certain games to be removed from said games.

Priceless

 

[drunkenmaster]

Shame DM did not understand this in some of his older posts.

You realise that he's trying to give a round up of what I was saying... then you agree and then say it's a shame I didn't understand.

This 100%

While physically possible unlike what Fud was saying, no company will design a card where one version has 512Gb/s, another version has 1TB/s of bandwidth and they have the same amount of raw horsepower. It's inefficient and wasteful to have far more bandwidth than the architecture can handle. The cost of having a second layout, second tape out, second design team is massive. While it is possible, the chances of any company doing this is lets say 0.02% and the chances of a company without bag loads of cash spending probably 20-30% more on a single card by having a second version like AMD doing it is even lower. It would be bad engineering, it's that simple. The only realistic possibility of same cards with two different densities is as with all existing cards I can think of, memory chips of different capacity being used.

 

[Tonester0011]

Older posts, the ones where you were telling everyone how wrong they are hoping its an 8GB card and how only 4GB is possible, your memory is lacking it seems

 

[africanos23]

Just a small little question ...

If it's 8GB HBM are we talking 8x1024= 8192-bit ?

 

[Kaapstad]

You realise that he's trying to give a round up of what I was saying... then you agree and then say it's a shame I didn't understand.



While physically possible unlike what Fud was saying, no company will design a card where one version has 512Gb/s, another version has 1TB/s of bandwidth and they have the same amount of raw horsepower. It's inefficient and wasteful to have far more bandwidth than the architecture can handle. The cost of having a second layout, second tape out, second design team is massive. While it is possible, the chances of any company doing this is lets say 0.02% and the chances of a company without bag loads of cash spending probably 20-30% more on a single card by having a second version like AMD doing it is even lower. It would be bad engineering, it's that simple. The only realistic possibility of same cards with two different densities is as with all existing cards I can think of, memory chips of different capacity being used.

More backtracking.

Go on admit you got it wrong.

 

[RavenXXX2]

Functionality spelled functionallity on that slide, is this a spelling error AMD would make?

Well spotted, I call fake.

 

[drunkenmaster]

Older posts, the ones where you were telling everyone how wrong they are hoping its an 8GB card and how only 4GB is possible, your memory is lacking it seems

More backtracking.

Go on admit you got it wrong.

I said I DON'T CARE if it's 4GB and I can see why if 2GB stacks aren't available it would be likely to be 4GB only but I was also the only person saying 8GB is more than possible with 2GB stacks and was the only person explaining why the HBM 2.0 spec meant nothing in regards to when 2GB stacks would be available.

 

[drunkenmaster]

http://forums.overclockers.co.uk/showpost.php?p=26981931&postcount=7

30th September last year, thread on HBM, my take on it. This has always been my stance regardless of what people are now claiming. This is what I've always said, 4 stacks, 2GB stacks will almost certainly come before HBM 2.0 is available, HBM 2.0 isn't required to get any individual increase in spec of HBM.

I was raising the possibility of 2GB stacks before anyone else, while everyone else was insisting it wasn't possible. Not the other way around.

....

Remember those are generations with big changes, the density is set to increase within a generation. IE the first gen chips will have 1GB stacks and almost certainly 2GB stacks very soon, but still at first gen speeds. We won't see 8 hi stacks or 2Ghz speeds on the chips till the second generation. When we get 2GB per stack in the first generation is unknown at this point.

I'm not at all convinced that the 390X will come with HBM yet and isn't just a rumour, but we are very close to HBM being financially viable so it's possible.

So lets say the 390X has a 512bit bus, it can have 4 stacks of HBM, each stack will be 1GB(potentially 2GB), each stack gives 128GB/s, meaning 4 or 8GB with 512GB/s.

It might be a 384bit bus with 3 stacks(3 or 6GB) giving 384GB/s of bandwidth. It will be more comparable to Tonga bandwidth efficiency than Hawaii but HBM won't be directly comparable, the other slides talk about effectively being able to send more commands at a time to a stack and thus have it wasting less time waiting to execute the next command, so it will be more efficient per GB/s than we currently have.

There is also a sensible limit to how many stacks can be used, each extra chip reduces yields significantly so putting 8 stacks on at 1GB a stack is just not going to happen any time soon. I'd say best case scenario is 4 stacks 2GB a stack. I wouldn't bet against seeing 6GB in 3 stacks on a 390X.

This has been my stance from day one, if there are only 1GB stacks available I think it will be 4 or 6GB, 8GB is more than possible but yields and costs of adding 9 chips on an interposer will hit costs significantly. I believe 2GB would come out sooner than everyone else thought and that if it was available we'd get 4 stack 8GB models basically when that memory is available.