More AMDGPU Patches For Linux 5.12 Point To First Sign Towards PCIe 5.0, FreeSync HDMI

[_r00t-]

Great. I'd like to see fixes about the ring gfx timeout/kiq issues (navi10) for the next kernel release too.

[polarathene]

If PCIe 4.0 made motherboard manufacturers upset to the point they had to put in a fan on the bus, then what will they put for PCIe 5.0 motherboards?

Water cooling block?

New Noctua partnership and product

[pal666]

If PCIe 4.0 made motherboard manufacturers upset to the point they had to put in a fan on the bus

it was not pcie 4.0, just one particular chipset. next chipset will be better.

[aufkrawall]

Great. I'd like to see fixes about the ring gfx timeout/kiq issues (navi10) for the next kernel release too.

When there is no way to reproduce or viable debug data available, I wouldn't expect much. E.g. I have never seen spontaneous instability with Navi cards since I've got one since November (+ a 6800 in a different system) and not before with Polaris either.

[Paradigm Shifter]

If PCIe 4.0 made motherboard manufacturers upset to the point they had to put in a fan on the bus, then what will they put for PCIe 5.0 motherboards?

Water cooling block?

Phase change cascade.

Alternatively, the entire motherboard becomes one big heatsink? (Think Asus TUF, but metal and actually useful...)

[Spacefish]

If PCIe 4.0 made motherboard manufacturers upset to the point they had to put in a fan on the bus, then what will they put for PCIe 5.0 motherboards?

Water cooling block?

This is not really the PCIe 4.0, but the chipset made by global foundries.. A almost identical (or identical) chip is used as the I/O Die in the Matisse CPUs (Ryzen 3000 / Zen2).
The chip is a real power hog, probably the CPUs could have a way better energy efficiency if they did not use the Global Foundries I/O die..
Some people guess, that AMD had contracts with GF, so they had to buy some silicon from them.. This is where the x570 and the Matisse I/O die comes from..
If you compare it to Renoir (Zen 2 in 7nm combined with I/O and GPU on a single TSMC 7nm die) you can see the power difference..

Compare the 4900HS to the 3700(X) for example, they are almost identical performance wise as they feature the same Zen 2 cores, however the Renoir chip uses ~15W less power. Renoir has 8MB of L3 instead of 32MB (Matisse), that´s why Renoir is slightly slower in some benchmarks.

If availiable i would have used a Renoir based APU in my PC instead of a Matisse CPU. However the Renoir based APUs are not really availiable to the consumer market, i bought two of them via grey channels, but officially they are OEM only.. I guess the reason behind this: They need a larger 7nm chip, and production throuput booked by AMD at TSMC is limited.. So they try to use their wafers aloted to them as optimally as possible.
Otherwise the 7nm APU is really really good for an office PC.. Low power usage, integrated GPU and a lot of computing power.
I would only use Zen 2 APUs for Office PC builds, if they where availiable in quantity.

Hope the supply situation improves mid 2021.. It´s an industry wide problem.. TSMC greatly underestimated the demand and Corona did it´s part as well.
TSMC could probably easily sell double the capacity in 7nm right now, if they had the production lines for it.

[pieman]

Great. I'd like to see fixes about the ring gfx timeout/kiq issues (navi10) for the next kernel release too.

Have you tried forcing pci-gen 3 in bios? Seems like most users with this issue are ryzen users running their cards in gen 4. Gen 3 appears to fix the problem. One user here in another thread had the same issue but after taking my advice switching to gen 3, the problem went away.

I've seen some nvidia ampere users have problems running their cards in gen 4 mode too.

[drSeehas]

This is not really the PCIe 4.0, but the chipset ...
If you compare it to Renoir (Zen 2 in 7nm combined with I/O and GPU on a single TSMC 7nm die) you can see the power difference..

Renoir is PCIe 3 not PCIe 4 (Matisse).

Compare the 4900HS to the 3700(X) ...

This compares apples and oranges.

[_r00t-]

Have you tried forcing pci-gen 3 in bios? Seems like most users with this issue are ryzen users running their cards in gen 4. Gen 3 appears to fix the problem. One user here in another thread had the same issue but after taking my advice switching to gen 3, the problem went away.

I've seen some nvidia ampere users have problems running their cards in gen 4 mode too.

I didn't mention that my motherboard supports pci-gen3 (x470 chipset) . My apologizes. This bug appears randomly when i'm running DooM Eternal(via Steam).

[verude]

WTF is that, why aren't they doing it for the standardized version found in HDMI 2.1+ ?
Or is it for the versions of HDMI prior to 2.1 ?

Blame the hdmi forum https://www.phoronix.com/scan.php?pa...pec-Hurts-Open

People should try to get devices with DP if they can so they aren't held hostage by them . (There are actually even a few tv devices with DP ports, but you'll have to specifically look for them as they are a minority among tv models).