Intel kills 10nm ?? oO

[smilingcrow]

So seemingly Samsung have no recent history of producing high wattage and high performance silicon for CPUs or GPUs and the GTX 1050 doesn't count in this context.

 

[Rroff]

So seemingly Samsung have no recent history of producing high wattage and high performance silicon for CPUs or GPUs

High performance might be pushing it but Vega 56/64 are produced on what is originally Samsung's 14LPP node - I wouldn't write off their capabilities at producing complex, high performance GPUs.

 

[Radox-0]

I dont believe for one min that Nvidia could just "drop" a 7nm chip at a moments notice, if they could they surely would have already.

While I doubt they could "drop" it at a moment notice, even if they could there is no need for them to do anything right now really, would be competing with themselves for most the part. Keep on with Turing, possibly make some refinements (such as upping memory speed if that proves to be true, like some Pascal SKU's), drop prices if need be and let that sell. Already have a glut of inventory in the market according to there market forecast on Thursday which is only now going to start easing with gaming chips: https://www.bloomberg.com/news/arti...ecast-signals-gaming-chip-inventory-is-easing no point putting anything else on line (not that they likely have anything ready). From there perspective given the current competition, or largely lack of in most the arena, keep back any easy improvements for next gen. Let Turing saturate the market, people have proven they are willing time and time again to upgrade every generation (and in some cases within the same generation).

 

[smilingcrow]

High performance might be pushing it but Vega 56/64 are produced on what is originally Samsung's 14LPP node - I wouldn't write off their capabilities at producing complex, high performance GPUs.

If they weren't actually produced by Samsung then that's a no then.
I'm not writing them off but curious if they have actually fabricated a high performance and high wattage part yet.

 

[4K8KW10]

So seemingly Samsung have no recent history of producing high wattage and high performance silicon for CPUs or GPUs and the GTX 1050 doesn't count in this context.

If they weren't actually produced by Samsung then that's a no then.
I'm not writing them off but curious if they have actually fabricated a high performance and high wattage part yet.

I'm out of the loop.

Enough said.

 

[pmc25]

I correctly ignored what you've said. Since you don't even mention that nvidia will probably use Samsung's 7nm process and not TSMC's.

Meaning you ignorantly ignored it. Samsung 7nm EUV (LPP) uses fewer layers of EUV than TSMC 7nm+ AFAIK. Since the latter uses 4, and the former does it for <20% of layers, total number of which is below 20. Samsung is also a larger pitch, and generally considered a lower power rather than higher performance solution than TSMC's 7nm+. Everything else being equal, Samsung 7nm LPP (EUV) will not perform as well as TSMC 7nm+ (EUV). But of course either could botch it, and Samsung will have more wafers sooner. Albeit as with TSMC, they will remain unavailable for a long time after launch to anyone but the mobile chip makers and their colossal order books.

Also, unless AMD have sewn up so much of the scraps left over from the mobile industry that TSMC are not tenable for NVIDIA's next major round of releases, I highly doubt they'll go Samsung for the majority. Not only would it potentially be seen as a risk by investors, and NVIDIA's share price is now under significant pressure, but Huang is incredibly close to the TSMC brass. I think they would have to be pushed, rather than jumping.

Furthermore, you're now moving the goalposts from DUV to EUV.

 

[4K8KW10]

Meaning you ignorantly ignored it. Samsung 7nm EUV (LPP) uses fewer layers of EUV than TSMC 7nm+ AFAIK. Since the latter uses 4, and the former does it for <20% of layers, total number of which is below 20. Samsung is also a larger pitch, and generally considered a lower power rather than higher performance solution than TSMC's 7nm+. Everything else being equal, Samsung 7nm LPP (EUV) will not perform as well as TSMC 7nm+ (EUV). But of course either could botch it, and Samsung will have more wafers sooner. Albeit as with TSMC, they will remain unavailable for a long time after launch to anyone but the mobile chip makers and their colossal order books.

Also, unless AMD have sewn up so much of the scraps left over from the mobile industry that TSMC are not tenable for NVIDIA's next major round of releases, I highly doubt they'll go Samsung for the majority. Not only would it potentially be seen as a risk by investors, and NVIDIA's share price is now under significant pressure, but Huang is incredibly close to the TSMC brass. I think they would have to be pushed, rather than jumping.

Furthermore, you're now moving the goalposts from DUV to EUV.

The problem is that you are driving the discussion in a direction which I don't like. I stated some figures - if you don't agree with them, just say no and say how much you expect. There is no need at all to describe the processes in details.

My expectation is that nvidia has a 150 sq.mm 7nm chip that is as fast as Vega 20.
And that it has a 350 sq.mm 7nm that is at least 50% faster than Vega 20.

What do you expect again?

 

[Rroff]

Also, unless AMD have sewn up so much of the scraps left over from the mobile industry that TSMC are not tenable for NVIDIA's next major round of releases, I highly doubt they'll go Samsung for the majority. Not only would it potentially be seen as a risk by investors, and NVIDIA's share price is now under significant pressure, but Huang is incredibly close to the TSMC brass. I think they would have to be pushed, rather than jumping.

Last I heard nVidia was still evaluating things like supply and unit cost and TSMC was far from a done deal. I don't have any newer information at the moment.

 

[AthlonXP1800]

Intel Xeon Roadmap Leaked Out, Unveils 10nm Ice Lake-SP With PCIe Gen 4 & Up To 26 Cores in 2020, Next-Gen Sapphire Rapids With PCIe Gen 5 & DDR5 in 2021

https://wccftech.com/intel-xeon-roadmap-leak-10nm-ice-lake-sp-2020-sapphire-rapids-sp-2021/

Good to see Ice Lake will have PCI Express 4.0 support to compete with Ryzen 3000 series but it seem PCI Express 4.0 will have shortest life in history compared to past standards and it will replace with PCI Express 5.0 12 months later. Look like a massive waste of time and money to implemented PCI Express 4.0 in CPUs, AMD and Intel should had skipped PCI Express 4.0 and straight to PCI Express 5.0 instead.

 

[4K8KW10]

Ice Lake will have PCI Express 4.0 support to compete with Ryzen 3000 series

It won't compete with Ryzen. It's xeon that goes against EPYC.

 

[LePhuronn]

AMD and Intel should had skipped PCI Express 4.0 and straight to PCI Express 5.0 instead.

But at least all AMD has to do is change the I/O die

 

[4K8KW10]

But at least all AMD has to do is change the I/O die

And then 20 years more without any change. No, thanks. Better stay with PCIe 4 for 5 years, and then move to PCIe 5, if devices which support it will have arrived at all.

 

[smilingcrow]

But at least all AMD has to do is change the I/O die

Aren't you forgetting the PCIe controller on the CPU which is used for a GPU and a primary SSD?

 

[4K8KW10]

Aren't you forgetting the PCIe controller on the CPU which is used for a GPU and a primary SSD?

That's on the IO die, not on the chiplet.

 

[LePhuronn]

Aren't you forgetting the PCIe controller on the CPU which is used for a GPU and a primary SSD?

No, it's on the I/O die in the CPU package.

 

[LePhuronn]

And then 20 years more without any change. No, thanks. Better stay with PCIe 4 for 5 years, and then move to PCIe 5, if devices which support it will have arrived at all.

lolwhut? You are loving the hyperbole these days, aren't you.

 

[4K8KW10]

lolwhut? You are loving the hyperbole these days, aren't you.

You don't agree that if PCIe 5 is released now, it will have zero use since no graphics cards, nor any other devices even support PCIe 4?

 

[smilingcrow]

No, it's on the I/O die in the CPU package.

Of course!
So they still need to significantly crank up the bandwidth between the cores chiplet(s) and the I/O die.
Wonder what headroom they have with the current design?

 

[4K8KW10]

Of course!
So they still need to significantly crank up the bandwidth between the cores chiplet(s) and the I/O die.
Wonder what headroom they have with the current design?

No. You can't even saturate the current PCIe 3 lanes, you can only dream of saturating double that with PCIe 4, and it's light years years away to even think about PCIe 5.
Just no.

 

[Grim5]

Well that confirms we're stuck with DDR4 for a couple more years at the very least