AMD Ryzen Z1 Extreme Proves Very Versatile For Power/Performance - Benchmarks Against The Ryzen 7 7840U

[phoronix]

Phoronix: AMD Ryzen Z1 Extreme Proves Very Versatile For Power/Performance - Benchmarks Against The Ryzen 7 7840U

With the ASUS ROG Ally being the first device powered by AMD's new Z1 Extreme SoC with Zen 4 CPU cores and RDNA3 graphics, it's been very interesting to see its performance advantages over the Steam Deck. But beyond its potential for use in gaming handhelds, it's quite fascinating to see how powerful the Z1 Extreme actually is when removing power restrictions on this SoC. In this article is a wide range of CPU benchmarks putting the Z1 Extreme up against the new Ryzen 7 7840U laptop SoC as well as prior generation Ryzen 7 PRO 6850U for reference. When adjusting the ACPI Platform Profile configuration, the Ryzen Z1 Extreme proves to be very robust from a low-power SoC delivering good battery performance up through pulling 50+ Watts while outperforming the 7840U.

https://www.phoronix.com/review/amd-z1-extreme

[fahrenheit]

Interesting results.. Michael would it be possible to rerun the Ally tests without ACPI profiles enable and using the same mechanism as both the Acer and Lenovo machine? The power usage charts are strange, on one hand the default profile appears totally stuck on 15w, the saver one hard limits to 15w with an average on 9w and the performance one, due to the excellent cooling solution, reaches absurd power usage values.
So much that by default I think the 6850U might provide better value than the Z1 Extreme..

[tildearrow]

Typos:

In the default mode of operation (balanced APCI Platform Profile), the Z1 Extreme

On a performnace-per-Watt basis, the power-saver mode was looking very good.

The Z1 Extrmee would be an interesting candidate for some thin clients

[Michael]

Typos:

Thanks!

[panikal]

The Z1 Extreme was consistently showing it's capable of outperforming the Ryzen 7 7840U found within an Acer laptop albeit at the cost of driving the SoC power consumption up to 50 Watts.

Michael how hot did the Ally get? I'd be scared of burning the thing up. Was the device able to be handled without it being uncomfortable?

Its not a normal realistic scenario for gaming, but that is a serious amount of *easily portable and cheap* off-grid compute for any potential radio/hacker needs...at least for an hour. I presume the Ally has USB-C charging and could take a 20kw backup to extend the time but I guess at running hard at 50w you start to reduce its future lifespan. Unless it managed a cool 30C under the hand?

[abufrejoval]

My main takeaway is that the Ryzen 7000 APUs are rather good chips, which can be tuned to fit distinct performance profiles significantly, if the platform and BIOS is open for that.

I'd just love to have them on a Mini-ITX board with open DIMM-Slots, so I can put 64, 96 or even 128GB of RAM, preferably ECC in them.
ECC was physically included on APUs, but unfortunately typically only enabled on certain SKUs.

I'd love using them for edge server HCI clusters, where power might not just come from the wall, heat dissipation ability or noise tolerance vs compute power demands are variable: if the base design allows such high degree of flexibility, I also want to use it.

I have a budget Mini-ITX board from Erying with an Alder Lake i7-12700H here, 6P and 8E cores which allows setting PL1 and PL2 at anywhere between 5 and 120 Watts on the fly and lets you control P vs E usage and core counts via numactl and it's quite fascinating to see what you can squeeze from 10 Watts or just how powerful it gets at 120 Watts (that mobile chip is thrashing my Ryzen 9 5950X at single threaded workloads!)

Overall the Ryzens should be rather similar in top performance, but more efficient overall, although when it comes to getting top performance out of say 10 Watts, I'm not sure I'd actually bet for a Ryzen win against the E-cores. The only thing I know for certain after testing is that the P-cores become dog slow at those power levels.

But my 5800U isn't exactly screaming when my notebook is tied to 15 Watts: at 28 Watts it becomes quite a different machine.

[brucethemoose]

My main takeaway is that the Ryzen 7000 APUs are rather good chips, which can be tuned to fit distinct performance profiles significantly, if the platform and BIOS is open for that.

I'd just love to have them on a Mini-ITX board with open DIMM-Slots, so I can put 64, 96 or even 128GB of RAM, preferably ECC in them.
ECC was physically included on APUs, but unfortunately typically only enabled on certain SKUs.

I'd love using them for edge server HCI clusters, where power might not just come from the wall, heat dissipation ability or noise tolerance vs compute power demands are variable: if the base design allows such high degree of flexibility, I also want to use it.

I have a budget Mini-ITX board from Erying with an Alder Lake i7-12700H here, 6P and 8E cores which allows setting PL1 and PL2 at anywhere between 5 and 120 Watts on the fly and lets you control P vs E usage and core counts via numactl and it's quite fascinating to see what you can squeeze from 10 Watts or just how powerful it gets at 120 Watts (that mobile chip is thrashing my Ryzen 9 5950X at single threaded workloads!)

Overall the Ryzens should be rather similar in top performance, but more efficient overall, although when it comes to getting top performance out of say 10 Watts, I'm not sure I'd actually bet for a Ryzen win against the E-cores. The only thing I know for certain after testing is that the P-cores become dog slow at those power levels.

But my 5800U isn't exactly screaming when my notebook is tied to 15 Watts: at 28 Watts it becomes quite a different machine.

Some benchmarker in a Discord server found that the 5800HS had higher task energy than the 4800HS, even though it was faster... But I can't remember specifics or find the post. Discord is an information black hole :/

[Teggs]

'Damn the wattage, full speed ahead!' - Asus


Asus does seem to be paying close attention to cooler design, even using liquid metal in some laptops for a while now, so credit to them. I don't take this as an indicator to competitors to run handhelds' APUs at 50W, but it is interesting that the ceiling is that high.

[darkbasic]

Why don't they make laptops with this thing then?

[ktecho]

How can a smaller device have more performance than a bigger device with better cooling and the same CPU?