AMD: R8xx Speculation
- Thread starter Shtal
- Start date
Like what? DX-Compute speeding up HDAO-computations by a factor of up to x.x times?
~2x ALUs up from RV740, which is also in 40nm plus DX11 with it's larger requirements of caches, need for Double Precision and HDR-compression for - presumably - the TMUs?
That'll leave plenty of room for the provisions of native dualcores, yes.
Something like this?
That was my idea anyway. Especially considering rumours about packaging costs being quite high and the process being quite difficult.
Maybe its something like RV840 -> Memory interface (FSB) <- RV840???
Ya, but the rumours about excessive packaging cost and difficulty may indicate a MCM type approach. The idea that its a native dual core is just a rumour at this point as well, but I would say the case for the MCM at this point is stronger.
Exactly!
I don't think the term "dual-core" could be appropriate for any MCM GPU solution -- more like "dual-die" or sort of.
Dual core makes it easier to understand for the mainstream reader IMHLO; if you say dual die they'll think of two separate cores which is not the case. Make it two dies on one package if that helps
Well semantics aside how you want to call the config in question, it's the question I asked further above. Assuming the higher packaging cost is real and there's an intermediate 1200SPs why not go from there to 1600SPs instead of MCM/dual core/dual die or whatever one wants to call it?
If the hypothetical 800SP chip is roughly in the 120-130mm2 league (as I'd personally estimate it) then a twice as complex 1600SP chip would be defintely in the =/>260mm2 range which is anything but big considering RV770 ranged roughly in that ballpark under 55nm.
Dual core makes it easier to understand for the mainstream reader IMHLO; if you say dual die they'll think of two separate cores which is not the case. Make it two dies on one package if that helps
On the contrary, it makes no sense whatsoever. If the "mainstream reader" as you wish to call them needs educating then they need to be told it the way it is. Dual-core is more confusing because the dual-core chips that your stupid "mainstream reader" is used to are single die. Spin like you're talking about just feeds the PR spin crap machine that'll inevitably start to orbit any MCM product if and when it emerges.
You're calculating the 800SPs with hypothetical 10 clusters in mind aren't you? Try with less clusters (and think of a good amount of less TMUs/ROPs which eat far more die area than ALUs) and it might work. I'm tapping in the dark here of course as anyone else.
I see your point and I won't say that you're not right. On a more humorus note as I've been supposedly attached to a certain other PR spin crap in the past, I guest it's time for AMD to host me under its wings
(and no for God's sake don't anyone take that ridiculous joke seriously...)ATI's current high end solutions are dual-GPU, MCPs. An MCP could in theory host 3 or 4 GPUs (just like what IBM did with the POWER4).
I just don't see how you'd call it a 'dual core'. That's about the most retarded thing in the world and even more confusing.
You already have the idiots in marketing for ATI and NV calling each FUNCTIONAL unit a core. Let's not encourage them with more nonsense. Let's also try not to overload terminology so badly.
DK
I just don't see how you'd call it a 'dual core'. That's about the most retarded thing in the world and even more confusing.
You already have the idiots in marketing for ATI and NV calling each FUNCTIONAL unit a core. Let's not encourage them with more nonsense. Let's also try not to overload terminology so badly.
DK
Dual core makes it easier to understand for the mainstream reader IMHLO; if you say dual die they'll think of two separate cores which is not the case. Make it two dies on one package if that helps
Well semantics aside how you want to call the config in question, it's the question I asked further above. Assuming the higher packaging cost is real and there's an intermediate 1200SPs why not go from there to 1600SPs instead of MCM/dual core/dual die or whatever one wants to call it?
If the hypothetical 800SP chip is roughly in the 120-130mm2 league (as I'd personally estimate it) then a twice as complex 1600SP chip would be defintely in the =/>260mm2 range which is anything but big considering RV770 ranged roughly in that ballpark under 55nm.
Assuming that theres a higher end SKU, then assuming that it won't be the one used for their high end model, isn't that an assumption upon an assumption?
Even knowing that a dual chip version is forthcoming, we still don't know how that would fit into their naming scheme. Is a 1600SP dual chip the 70 model or is it the 70 X2 model? Even if its the X2 model, does that mean that there may also be an X4 model? Its difficult to piece together what will actually happen here.
My best guess is this, they will try to get the most mileage out of a single chip as they can. So RV840 means single chip, RV870 is double RV840 and R700 is quadrupel or something to that effect. RV840 is the 170-80mm^2 version. So the 5850 is single chip, 5870 is double and what we would call the 5870 X2 is quad RV840. RV 830 is probably going to be a mainstream 56xx part IMO.
However I remember a tweet from that Catalyst guy and he implied that the current naming scheme is going to be changed, I guess if they are going to go multigpu/same die/module they really do need to distinguish the difference.
All this talk abput multichips requires that they've made some huge breakthrough on multichip rendering and/or not using AFR, since with current tech every added chip increases the performance less and less, there's just no point in trying to use for example quad chips on a single card, not to mention that a mere 1GB model would need to have 4GB mem on board
Similar threads
