Baseless Next Generation Rumors with no Technical Merits [pre E3 2019] *spawn*

Status
Not open for further replies.
If you can wait until I return to work, I can provide one for the PS5. Hope you don't care for a little bad english, and consider it instead the proof of authenticity.
 
I think his argument is why would you use two cpu chiplets instead of one? You are basically using an entire chiplet for nothing more than cooling purposes. Is that the cheapest way to combat a cooling issue?

Exactly and I don't think cooling those 8 cores is an issue at all.

A CPU chiplet and a separate GPU makes sense, but two CPU dies with only half activated is a huge waste, very much in opposition what a console chip design is.

I guess if they could use one of those packages fully enabled in a server/cloud instances and get economies of scale that way it could make sense to have that type of design, but even then that's a whole lot of waste in the consumer product.
 
I think his argument is why would you use two cpu chiplets instead of one? You are basically using an entire chiplet for nothing other than cooling purposes. Is that the cheapest way to combat a cooling issue?

As I said it could be simply cheaper because AMD can sell them damaged chips which are probably difficult to sell in low end 4core Ryzen3 if they even plan such products.
 
As I said it could be simply cheaper because AMD can sell them damaged chips which are probably difficult to sell in low end 4core Ryzen3 if they even plan such products.

That is a lot of damaged chips then considering the volumes an Xbox console is going to have. AMD has a lot of avenues to use such possible chips like a 12 core Ryzen variant using two chiplets. I also don't think 8 core is the lowest end from now on. Realistically there is no way using two CPU chiplets is cheaper for MS than using one of the same kind.
 
So what if these damaged chips don't reach the base clock for their normal product line? Throw them away? The list price for a 8/16 Zen at obviously higher base clock is 178USD. Even if that's not the real price MS would pay why should AMD give them discounts for fully functional chips they could sell at full price?
 
Where does it say two cut down CPUs to make one?

I see to RCC that are gated to 50% and using two, but for CPU the dual CPU is for Azure

I assume that is for running two Lockhart 1080p streams from a single GPU timesliced and each getting it's own CPU. It also mean when not streaming it's more potent for Azure workloads.
 
So what if these damaged chips don't reach the base clock for their normal product line? Throw them away? The list price for a 8/16 Zen at obviously higher base clock is 178USD. Even if that's not the real price MS would pay why should AMD give them discounts for fully functional chips they could sell at full price?

You assume that these defected chips that don't fit any other use are a meaningful number out of the total, I don't believe that to be the case. Some chips will go to trash, that's the nature of silicon manufacturing. I don't assume MS getting highest bins of these chips, but considering the clock speeds they are operating, they don't have to.

To be honest I personally still believe a single APU-design is the most likely scenario for the next gen consoles. I don't see multichip solutions bringing concrete benefits for the console makers. AMDs CPU business has other reasons to use chiplets that the console makers don't have. I could be wrong though of course.



Where does it say two cut down CPUs to make one?

That hand written thing "clearly" stated so. It says that the reason for 2 partially actived chiplets instead of 1 a fully active is mainly due to cooling.
 
Last edited:
If you can wait until I return to work, I can provide one for the PS5. Hope you don't care for a little bad english, and consider it instead the proof of authenticity.

Yes, please.

[...]

I guess if they could use one of those packages fully enabled in a server/cloud instances and get economies of scale that way it could make sense to have that type of design, but even then that's a whole lot of waste in the consumer product.

Looking at the specs and page 5 picture that seems actually what the "source/leaker/faker/whatever" was implying for xCloud. The server hardware would be the same as Anaconda but instead of two cut down Zen 2 chiplets it would use two 8 core Zen 2 chiplets to run two Lockhart streams.

But two chiplets instead of one should introduce some latency, right? Also, considering AMD already has a rumored yield of ~70% [0][1] for Zen 2 chiplets would using two defective chiplets really offset the added packaging complexity?

[0] https://www.overclock3d.net/news/cp...uVrfanB-mziRJSJIldcTrB7PwAXx0u2oXyxo9K8BhtICY

[1] I actually would have expected them to have better yields since the chiplets are mobile SoC sized, but I may have underestimated quad patterning etc.

Where does it say two cut down CPUs to make one?

[...]

From what I see it's only implied by "2 Zen 2 chiplets 8c/16t". Since we know Zen 2 chiplets have 8 cores that would mean they use two cut down or defective chiplets which only have 4 cores each.
 
Surprised many of you can read it at all. It drives me nuts that their persons “ look like 11. And their ‘1’s look like ^. The handwriting is marginally better than doctor prescriptions.

Though thanks for sharing this is a good place to discuss. While I agree with your friend it’s not worth trying to reverse engineer to validate, there are some concepts in here I haven’t seen in previous “next generation” guesses, which seem to be sparking some interesting discussion at least in the idea that perhaps there are more configurations we should consider
 
Page 'Layout', right side:

Not enough functionning CUs ? Use in Lockhart by deactivating CUs and memory controllers

Based on this Lockhart GPU should have the same size as Anaconda GPU.

Page 'Layout', comparing left side and right side we clearly see than Lockhart GPU is about twice smaller than Anaconda GPU.
 
So am I looking at this correctly 5.587 TF for LockHart and 11.174 TF for Anaconda? This mofo writes 1s like when I fold my clean laundry, which is beyond unrecognizable. This also makes me think with the relatively "lowish" Tflops for Anaconda maybe MS is undercutting Sony by going $399 for the high end sku when the world thinks they want the most powerful console again. Or maybe our expectations are just too high.
 
Maybe the TFs dont matter as much if you have better RT-RTs? This time around the entire possibilities of next-gen is more interesting than the previous because of the unknown RayTracing feature-set.
 
FWIW, RA says it's not legit, but the one centered on the HBM patent is. I've asked for clarification regarding RCC, since they both mention it.
 
Last edited:
As for RCC, is that some sort of abbreviation for Raytracing? Like Raytracing Compute Core? Raytracing Central Core?
 
Status
Not open for further replies.
Back
Top