@bolded: We don't even know if that's a true chip (and at 20CU, I really doubt it, especially considering it will be totally bandwith starved even with DDR4 4000). But I digress.

The actual cost depends how much AMD has to pay per wafer, divided by how many chips on that wafer are salvageable for that purpose. So let's say a wafer cost 1000$ (I'm just making up a price here), 20 such chips would fit on it but only 10 would be fully functioning, the others would have to be sold as either 3400G or binned entirely due to defects. In this case AMD would certainly charge at least 100$ on the 3600G to cover the costs already, and use the 3400G for winnings.

However, on a console that's not possible, hence why the PS4 has 2 deactivated CU to improve the yield rate.

@italic: These costs are not always covered, I can remember that the cost of some chips were actually worked into the yearly contracts instead of receiving a sum early on. And considering AMD didn't seem to have gotten any Lump sum (if they did, it doesn't show up in the financial reports at least), I do think they have to cover for those expenses with the chip sales.

@underlined: Well, no, I'm not saying that they won't do it anymore, but rather that they are not obliged to do so anymore to have any sizable income at all.

At the time when the PS4/XBO came out, AMD CPUs were doing very badly and were the laughingstock of the industry. They just released Hawaii, but had much problems keeping up with NVidias updated Kepler (GeForce 700 series), so earnings were breaking away left and right and could only really compete over the price. As a result their profit margin plummeted, and still is awfully low for the sector (it's under 50% while Intel and NVidia are close to or above 70%; at the time it even dropped below 30%, which is bad in any sector). All this made that AMD was desperate for some stable income, which made Sony and Microsoft holding all the cards during the price negotiations. But that won't be the case this time, and AMD will squeeze some winnings out of the chips.

Also, as a side note, you give the costs at 30-40$. Tell me how that works if about half of the sales are from console chips (which was true in 2016) yet the profit margin is at only 24%? Do you think AMD sold their other chips all below production price? And how could that be, considering most chips cost much more than the one in the PS4? Or do you think they had such an R&D expense that it covers half the expenses before wages and taxes? Just saying that your price is off, it may be well below 100$ by then, but I don't think anywhere close to the numbers you're putting there, more like 60-80$. Don't forget that 350mm² ain't exactly a small chip (a 10 core Skylake-X is only 322mm², for instance) and that such a big chip normally sells at quite some higher prices for reasons detailed above.

Your Apple example is a bit special, they use it due to OGL and OCL capabilities, where NVidia is weaker than AMD and generally has been like that. Them being cheaper than NVidia is only icing on the cake. But that's going to change soon anyway, considering that Apple wants to design all their chips in-house and are migrating everything to ARM.

Perhaps, but developers do, and aiming for higher fps means better margins. As it stands, a lot of games with 30 fps aim dip into slideshow territory, games like AC: Odyssey are a good example, I don't think I've ever seen worse frame rates in a modern AAA title. With more effects and higher resolutions, severe drops in frame rate become even more jarring, a more advanced and crisp video or render stands out all the more when it slows down. 

Xbox One X doesn't provide proper 4K, it uses checkerboard rendering and rather few effects, frame rates on many titles are very low, Destiny 2 runs at a somewhat unstable 30 fps.

Half-assed resolutions with poor performance is not what developers want to work with, it's better to find an entry point where they can get both visual fidelity and performance and not be forced to choose. And if the hardware ends up too costly, developers and publishers have a smaller market to sell their software to. I'd much rather the standard be a stable 1440p with full effects and shading than stripped down or faked 4K, then perhaps release a more expensive version of the console that does more or less what the Pro and One X do to the line-up right now.

The fact that 30 fps is still more or less industry standard in console gaming in 2019 is downright shameful, especially since games often dip well below.

https://www.macrotrends.net/stocks/charts/AMD/amd/profit-margins

https://ycharts.com/companies/NVDA/profit_margin

https://ycharts.com/companies/INTC/gross_profit_margin

70% profit margin isn't something common at all.

@bolded: We don't even know if that's a true chip (and at 20CU, I really doubt it, especially considering it will be totally bandwith starved even with DDR4 4000). But I digress.

The actual cost depends how much AMD has to pay per wafer, divided by how many chips on that wafer are salvageable for that purpose. So let's say a wafer cost 1000$ (I'm just making up a price here), 20 such chips would fit on it but only 10 would be fully functioning, the others would have to be sold as either 3400G or binned entirely due to defects. In this case AMD would certainly charge at least 100$ on the 3600G to cover the costs already, and use the 3400G for winnings.

However, on a console that's not possible, hence why the PS4 has 2 deactivated CU to improve the yield rate.

@underlined: Well, no, I'm not saying that they won't do it anymore, but rather that they are not obliged to do so anymore to have any sizable income at all.

At the time when the PS4/XBO came out, AMD CPUs were doing very badly and were the laughingstock of the industry. They just released Hawaii, but had much problems keeping up with NVidias updated Kepler (GeForce 700 series), so earnings were breaking away left and right and could only really compete over the price. As a result their profit margin plummeted, and still is awfully low for the sector (it's under 50% while Intel and NVidia are close to or above 70%; at the time it even dropped below 30%, which is bad in any sector). All this made that AMD was desperate for some stable income, which made Sony and Microsoft holding all the cards during the price negotiations. But that won't be the case this time, and AMD will squeeze some winnings out of the chips.

Also, as a side note, you give the costs at 30-40$. Tell me how that works if about half of the sales are from console chips (which was true in 2016) yet the profit margin is at only 24%? Do you think AMD sold their other chips all below production price? And how could that be, considering most chips cost much more than the one in the PS4? Or do you think they had such an R&D expense that it covers half the expenses before wages and taxes? Just saying that your price is off, it may be well below 100$ by then, but I don't think anywhere close to the numbers you're putting there, more like 60-80$. Don't forget that 350mm² ain't exactly a small chip (a 10 core Skylake-X is only 322mm², for instance) and that such a big chip normally sells at quite some higher prices for reasons detailed above.

I said before taxes, you give me after taxes. Though I understand that I explained that only far down the text, so the error is on me, mea culpa.

With that knowledge, you'll see that the profit rate AMD had before taxes and operating costs (the gross profit margin, couldn't find the word before) is about what NVidia does percentage wise  after all operating costs have been taken into account. And that one dropped to 54% in the last quarter while having been above that for many quarters before. Same for Intel, who even mostly stay above 60%.

AMD, on the other hand, is mostly between 20 and 40%, with one quarter even dropping to just 4%. But the general trend is upwards, I don't doubt they can break into the 40% range this year. But for that they need to make some money from their chips.

 Now each wafer costs anywhere between $300 and $10000 on the foundry side of things depending on number of processing steps and complexity.

AMD will pay the agreed amount for every wafer..... regardless of what is working or not working in it as long as an agreed upon  minimum chip yield per wafer is met.

But we know the vega based 2400G exists. We also know that AMD always makes a few APUs with every product series. And those things aren't designed to be graphical powerhouses anyways so RAM bottleneck is a moot point.

And you have got this backwards, AMD even needs their money less now than it ever has. So its more likely to work with them for what is more or less a licensed chip than to tr and milk them for anything.

Again you are going about this wrong...... yes back in 2016 growth from semi custom sector (which is probably 90% consoles) equated to about half of AMDs quarterly revenue in certain quarters. That's total revenue not profit margins. There is a very big difference. eg in a particular quarter revenue in that division was at $590M (real number in their 2nd quarter 2016). Now if in that quarter alone they took in orders of say 5M chips and got around $100 for each one what does that give you? Yup.... around $500M.

Still hard to understand?

AMD gettig $100 for ach chip they give to sony isn't them selling them at bargain prices at all. Thats them selling them at "bulk/OEM" pricing which is totally normal when any company puts in orders in the region of millions. 

Take the 3600G or instance, say AMD sells that at retail for $220, that pansout like this... the actual cost of making each of those chips (what AMD pays to the foundry) is like $30/$40. Then AMD will add their markup to account for things like yields, profits, packaging and shipping..etc. At this point the chip comes up to around $170. Then they put their MSRP sticker price of $220 so te retailers make their own ut too.

@bolded: We don't even know if that's a true chip (and at 20CU, I really doubt it, especially considering it will be totally bandwith starved even with DDR4 4000). But I digress.

Specifying some of AMD's improvement is irrelevant as long as you don't also specify what nVidia has achieved. A lot of the engineering work goes into improving the performance and power efficiency by switching from third pary cell libraries to custom IC designs for a particular process node. Something nVidia obviously has spent a lot more resources on than AMD and that's something which doesn't show up as a new feature in marketing material.

I spend much of my working time analyzing GPU frame-traces, identifying bottlenecks and how to work around them. Every GPU architecture has bottlenecks, that's nothing new, it's just a matter of what kind of workload you throw at them. I have full access to all the performance counters of the GCN achitecture, both in a numerical and a visual form. For instance, I can see the number of wavefronts executing on each individual SIMD of each CU at any given time during the trace, the issue rate of VALU, SALU, VMEM, EXP, branch instructions, wait cycles due to accessing the K$ cache, exporting pixels or fetching instructions, stalls due to texture rate or texture memory accesses, number of read/write accesses to the color or depth caches, the number of drawn quads, the number of context rolls, the number of processed primitives and percentage of culled primitives, stalls in the rasterizer due to the SPI (Shader Processor Input) or the PA (Primitive Assembly), number of indices processed and reused by the VGT (Vertex Geometry Tessellator), the number of commands parsed/processed by the CGP/CPC, stalls in the CPG/CPC, number of L2 read/writes, L2 hit/miss rate. That's just a few of the available performace counters I've access to. In addition to that I have full documentation to the GCN architecture and I've developed several released games targeting it. Based on that I've a pretty good picture of the strengths/weaknesses of the architecture and I'm interested in hearing if you perhaps have some insight that I lack.

The geometry rate isn't really a bottleneck for GCN. Even if it was, the geometry processing parallelizes quite well and could be solved by increasing the number of VGTs. It won't be a problem in the future either for two reasons. 1) The pixel rate will always be the limiting factor. 2) Primitive/mesh shaders gives the graphics programmer the option to use the CU's compute power to process geometry.

I asked you to specify the inherent flaws and bottlenecks in the GCN architecture that you claim prevents the PS5 from using more than 64CUs, not AMD's marketing material about their GPUs. So again, can you please specify the "multitude of bottlenecks".

I'm not sure what you mean. It clearly says the area reduction is 70% and a 60% reduction in power consumtion. Pretty inline with what I wrote. An area reduction of 70% would yield a density increase of 3.3x. Probably just a rounding issue.

Here are the links to TSMC's own numbers.

https://www.tsmc.com/english/dedicatedFoundry/technology/10nm.htm

https://www.tsmc.com/english/dedicatedFoundry/technology/7nm.htm

DDR4 4000 can offer more bandwidth than HBM2. It is entirely how wide you wish to take things... But before then you reach a point where it's more economical to choose another technology anyway.

However... Considering that current Ryzen APU's with 38GB/s~ of bandwidth are certainly bandwidth starved with 11 CU's... I doubt that is going to change with 20CU APU's that have 68GB/s~ bandwidth.

But if you were to run that DDR4 4000 DRAM on a 512-bit bus, suddenly we are talking 256GB/s of bandwidth, which is more than sufficient for even a 40 CU count.

Knowing you, I'm sure you are aware that to connect the RAM sticks with a 512 bit bus would need many more layers on the motherboards, thus making them hugely expensive, so not exactly an economic solution. It's after all also the reason why we're still stuck with just dual channel, quad channel would help iGPU/APU a lot but make the boards much more expensive.

However, what I could see as possible would be reintroducing the sideport memory, in this case as a 2-4GB HBM stack, functioning as a LLC. But for that I would have expected special boards for APUs, like a 540G and 560GX, similar to the 760G/790GX during the HD 4000 series. Just with a very fast sideport this time, please.