AMD Athlon 3000G Linux Performance Benchmarks - The New $50 Processor

[nuetzel]

@Michael

Do Mesa (git) running on this realy cheap 3000G smooth?
Maybe you could do some A10-7850K (7860K) vs 3000G runs?

My fathers A10-7860K hangs all the time during boot at switching to 'amdgpudrmfb'.

Works only with 'nomodeset' ;-(((

[Michael]

@Michael

Do Mesa (git) running on this realy cheap 3000G smooth?
Maybe you could do some A10-7850K (7860K) vs 3000G runs?

My fathers A10-7860K hangs all the time during boot at switching to 'amdgpudrmfb'.

Works only with 'nomodeset' ;-(((

There are 7850K vs. 3000G graphics and CPU runs in this article, isn't there? (I don't have any 7860K)

[Tomin]

If it would've been your pick Tomin, would you go for a 3000G or a 240GE (both are APU's, right?) ?

It doesn't really matter. With so small differences, I'd pick whichever is cheaper.They are both fine CPUs. If CPU power matters then I'd think about four core models (Ryzen 3).

APU is AMDs name for a CPU that has intergrated GPU. ECC is a bit of a preference thing in these home servers, so choose whatever makes sense for you, financially or otherwise.

[Mez']

@Michael

Do Mesa (git) running on this realy cheap 3000G smooth?
Maybe you could do some A10-7850K (7860K) vs 3000G runs?

My fathers A10-7860K hangs all the time during boot at switching to 'amdgpudrmfb'.

Works only with 'nomodeset' ;-(((

I also have the A10-7860K, although it's coupled with a RX 560, hence I've barely used the GPU bit when I bought it. Can't talk much about the whole APU but the CPU has been serving me well.

Anyway, I'm thinking about upgrading my 3 years old rig to a Ryzen 5 2600 and AM4/DDR4 (FM2+ is DDR3), while keeping the RX 560 for now.
Could anyone tell me if that upgrade is sufficient to noticeably boost fps in games?
Or should I just keep the A10-7860K and go for a new GPU (RX 5500) instead? I mostly use it on a 4K60 65" TV to watch movies/shows (vdpau) and do occasional gaming and I would expect a minimum 50% improvement in fps (current x 1.5).

[Mez']

Some datapoints from my experience:

• Tomb Raider (2013) (1080p, most likely the Ultra quality setting):
  • A10-7850K: 103.3 FPS
  • Ryzen 5 1600: 146.6 FPS
• Middle-earth: Shadow of Mordor (1080p,custom quality setting):
  • A10-7850K: 36.23 FPS
  • Ryzen 5 1600: 61.24 FPS
• The above measurements were obtained with the same GPU that is about 2 times faster (in terms of FP32 TFLOPS) than RX 560.
• Memory bandwidth and latency:
  • A10-7850K (DDR3): about 20 GB/s and 80 ns
  • Ryzen 5 1600 (DDR4): about 40 GB/s and 90 ns

If the GPU is being underutilized (e.g: radeontop -d log.txt) during the task, then upgrading the CPU without upgrading the GPU will achieve higher framerates. Otherwise (that is: with an already saturated GPU) upgrading the CPU won't bring higher framerates.

Wow. That's exactly what I was looking for and couldn't find on a benchmark search. I guess the relative and extrapolated difference should be similar to these CPUs.
With my allocated budget, I would be very happy with 60 fps at 1080p high details (not ultra) in Dirt Rally, Kona, The Long Dark, Grid Autosport, Road Redemption and the likes.
I will look into the output of radeontop when I get home tomorrow, since I'm away for Christmas and all.

Huge thanks for this!

[willmore]

My take away was that the Ryzen 3 1200 at $60 is a great deal if you don't need graphics on chip.

[Adarion]

35w TDP ... Ryzen 3 2200GE ...

Indeed, it is a little sad that the G_E_ version is extremely hard to come by. And that it still lacks some Kabini-successor for the normal-user market. Something 25W TPD and below.
But I guess the Ryzen 2200G might still be an option. Powerful 4 real cores (and no possibly vulnerable SMT), but already coming with an acceptably low TDP that you can likely downtune + further downtune by downclocking and undervolting. The GE version also has lower stock clocks, plus probably selected DIEs to reliably run at lower voltages.
On the other hand, even something with more TDP towards the top doesn't mean it'll consume much during idle, thus it might still be suitable for file servers and the likes, if downclocked / power gated by the PM units/drivers during idle - and if you happen to need more bang it's readily available.

I'm still with my Kabinis for these purposes and they're still "good enough" (TM). If I need power I still have a FX and Ryzen 7 2700 (Zen+) (8C / 8 or 16T, 8x 3.2 turbo 4.1 at 65 W TDP) around. I'm seriously impressed that this beast compiles like mad and eats video transcoding like crazy and crunches numbers at mere 65W TDP.
Back in the days I also had a 4850e (well, still have ) at 45W TDP plus a passively cooled GPU. Tech has come a long way since then. And these days an APU is more efficient for lower wattages, just that you can't exchange things separately.

[edwaleni]

I just ran this test script against the Ryzen 5 3500U. I really like the new HTML output post test. Well done!

https://openbenchmarking.org/result/1912260-HU-1912244HU38

[tajjada]

So Michael ... you are comparing budget cpus whose most important selling point is the low price ... where are the perf/$ graphs?

Only including raw performance comparisons doesn't help people who want to choose the best value for money, which is what these cpus are all about.

[S.Pam]

Just got myself a 3000G and an Gigabyte B450 motherboard to replace my old E-350 setup. Will swap over it tonight. I guess I'll soon see if there is a big change in performance. At least for compiling stuff since I run Gentoo . Nice thing is that I can change to any Ryzen CPU on this MB in the future!

Update: So, after a few hours I can say that based on compile times the 3000G is 10x faster than the E-350. Not bad at all