AMD Athlon 200GE: Benchmarking The $60 Zen+Vega Chip

[dungeon]

I think it's just that compiling depends more on core count than thread count, ie multiple threads per core doesn't help as much.

That was off even back in february, Ryzen 3 1300X should be something like 2200G but results are continously way off in this benchmark for Ryzen 3 1200/1300X CPUs:



https://www.phoronix.com/scan.php?pa...idge-cpu&num=4

[S.Pam]

phoronix what is your opinion on this?

[dungeon]

Someone who have either Ryzen 3 1200 or 1300X should run this pts bench, might be that something is broken in Michael setup for these.

If not that, then could be bug in pts or even kernel bug for these CPUs.... whatever, obvoisly results continously goes way off then they should.

[bridgman]

That was off even back in february, Ryzen 3 1300X should be something like 2200G but results are continously way off in this benchmark for Ryzen 3 1200/1300X CPUs:

Whoa, I hadn't noticed that before. Agree those don't look right.

[grok]

Haswell Core i3 (i3 4130) takes quite a beating though less inconsistent.
I find this fishy :
4C/4T Haswell vs Kaby Lake i.e. i5 4670 vs i3 8100 are about the same. 2C/4T Haswell vs Kaby Lake i.e. i3 4130 vs i3 7100 is a massacre, which the 500MHz clock speed difference doesn't account for.

[msroadkill612]

The focus here is power usage, especially at idle.

It bears noting, that there are ways (dont ask me) of kinda using the magic zen+/apu auto tuning utilities (xfr2/pb2/pbo) in reverse - to underclock/undervolt etc. dynamically.

Any mavens out there have a view on this?

Our low idle power desires have to be balanced by our need for it to get the job done when busy.

The ability of this new dynamic cpu tuning is clear from zen+ overclocking reviews - ~"better than expert's efforts".

If it can be applied as suggested above - to dynamically downclock as priority - it seems a relevant consideration - maximising the power use ambit between load and idle, w/o sacrificing much top end performance.

[msroadkill612]

Well, might be that you have there something special mobo and PSU combo really , here for example that i5-4570 shows not 10W but 42W at idle:

http://www.benchmarks-tests.com/cpu/...r-Consumption/

Or maybe you have S model there - i5-4570S ? Googling around for that, people talk about 36W idle there (might depend also what you have connected there):

https://willhaley.com/blog/power-con...a-kill-a-watt/

I really can't find anyone else to talk of these about 10W idle for that model So i would guess as If you achieved that, you indeed have something special there

Are you running server without X or something like that maybe?

At the end if nothing special you did there and is goes so low as 10W, might be that you even have a bug Are these 10W idle on both using Linux and Windows?



But, who is in the range of 20W? It is more like about 25W. on average using Windows for AMD APUs and comparable Intel CPUs too, now if that same machines using average Linux do 30W idle there, that difference is nothing so much to be worried about

he is certainly derailing the discussion if 10w is not true.

Re cost (US 15c kw/h?) & cooling, even this worst case ambit of 20w (10w vs 30w) doesn't seem a big issue.

[R41N3R]

Well, might be that you have there something special mobo and PSU combo really , here for example that i5-4570 shows not 10W but 42W at idle:

http://www.benchmarks-tests.com/cpu/...r-Consumption/

Or maybe you have S model there - i5-4570S ? Googling around for that, people talk about 36W idle there (might depend also what you have connected there):

https://willhaley.com/blog/power-con...a-kill-a-watt/

I really can't find anyone else to talk of these about 10W idle for that model So i would guess as If you achieved that, you indeed have something special there

Are you running server without X or something like that maybe?

At the end if nothing special you did there and is goes so low as 10W, might be that you even have a bug Are these 10W idle on both using Linux and Windows?



But, who is in the range of 20W? It is more like about 25W. on average using Windows for AMD APUs and comparable Intel CPUs too, now if that same machines using average Linux do 30W idle there, that difference is nothing so much to be worried about

I build this machine quite long ago. The link you've provided regarding the i5 4570 might show increased power consumption due to the mainboard. For reference a old link to these 10 Watt systems, it has an Intel DH87RL mainboard https://www.heise.de/ct/artikel/Bauv...ite=4#anchor_7. Well, looking at the numbers I must say that the mainboard manufacturers should probably do better work then ;-) I've underestimated this. The Athlon 200GE is anyway quite interesting. And with Linux 4.20 at least the Ryzen 2*00G will additionally be able to turn off the graphic core, might save some power too.

[dungeon]

Intel DH87RL mainboard Well, looking at the numbers I must say that the mainboard manufacturers should probably do better work then ;-) I've underestimated this.

Mobos or PSUs as i said, some are more idle efficient some aren't.

[torsionbar28]

Mobos or PSUs as i said, some are more idle efficient some aren't.

At such low consumption figures, one would be better served with a DC-DC PSU like the PicoPSU, combined with a laptop power supply.