AMD Ryzen 5 8500G

[Wirko]

that they do have scheduling issues, or that it's just underwhelming.

All processors with heterogeneous cores, multiple CCDs and/or HT will have scheduling issues. (That leaves out only Sierra Forest ... haha?) The issues are just harder to detect and track if "strong" and "weak" cores are too closely related.

 

[oZ65]

16.1

Thanks, any chance of one of the AVX512 / VNNI256 / VNNI512 versions will be used for the next review?

It might be a good way to investigate the next Zen 5 architecture tbo.

 

[W1zzard]

Thanks, any chance of one of the AVX512 / VNNI256 / VNNI512 versions will be used for the next review?

It might be a good way to investigate the next Zen 5 architecture tbo.

The score reported is for the fastest version, tested for each CPU (x86-64-vnni512.exe for the 8500G)

 

[Scour]

I know that most/all testers use High-End-Mainboards, but especially for an APU it would be more realistic to use A/B-chipsets mainboards.

 

[W1zzard]

I know that most/all testers use High-End-Mainboards, but especially for an APU it would be more realistic to use A/B-chipsets mainboards.

That would make it really hard to compare scores because it'll be impossible to tell whether a result is due to the motherboard or due to the processor

 

[Scour]

Should an A620-mainboard really be slower with an APU than an X670-mainboard?

I think the typical APU-buyer prefers a not-High-End, power saving platform

 

[Vincero]

No it shouldn't, but it rules it out as a matter of course (STAPM type BIOS issues aside).
Also, in terms of consistency testing on the same platform for that and other CPUs going to the maximum, why switch boards?

 

[Kapone33]

I would argue that if reviewing APUs, TPU should be using As Rock boards. They support 120hz on the HDMI ports on some AM4 and most AM5 boards. Even some of the cheapest ones support that. That makes a huge difference for TVs with VRR and monitors with Freesync support. Spending $150 to get a proper As Rock board should not be too expensive for a review site. I know that this topic is not discussed in reviews but for me it is one of the things that make APUs so good. To get 120hz on a DGPU you need at least 6000 series from AMD. That makes a 6500XT feel smoother than a RX580 at 1080P.

 

[Scour]

Also, in terms of consistency testing on the same platform for that and other CPUs going to the maximum, why switch boards?

Consumers which use an APU won´t buy biggest motherboards with the most power consuming chipsets

Idle power consumption is a big topic for me and I think it´s more realistic with an A-chipset-board/low entry B-chipset-board for AMD

Otherwise, I could be interesting if Techpowerup also compare the power consumption differences of the tested mainboards in a graph for a quick overview

 

[AusWolf]

Consumers which use an APU won´t buy biggest motherboards with the most power consuming chipsets

Idle power consumption is a big topic for me and I think it´s more realistic with an A-chipset-board/low entry B-chipset-board for AMD

Otherwise, I could be interesting if Techpowerup also compare the power consumption differences of the tested mainboards in a graph for a quick overview

Let's be simple because you don't seem to be getting it...

If you test a 9950X in a high-end board, and an 8500G in an entry-level one, then can you say with 100% certainty that the 8500G is exactly X% slower only because of the chip itself, and the board has nothing to do with it? Of course you can't.

Testing consistency and assumptions about what a "typical buyer" would choose are entirely different worlds.

 

[Vincero]

Consumers which use an APU won´t buy biggest motherboards with the most power consuming chipsets

Idle power consumption is a big topic for me and I think it´s more realistic with an A-chipset-board/low entry B-chipset-board for AMD

Otherwise, I could be interesting if Techpowerup also compare the power consumption differences of the tested mainboards in a graph for a quick overview

So, what you're really after is actually a motherboard review, as even the chipset isn't the decider of power consumption - onboard features will have an impact, and the VRM component choices can easily eat up any efficiency benefits that might exist between an A/B/X chipset board.
Hardware Unboxed regularly do motherboard roundups covering performance and efficiency, and many a year ago so did sites like Anandtech and Tomshardware (something that seems to be less common now from them).

Only @W1zzard & co can really decide if / how to go down that route. One thing I would say though is that one change they could make if doing such a thing as an X670 board roundup, is to throw in a very common A and B series board for comparison so that cross comparisons might be easier.

 

[Scour]

Let's be simple because you don't seem to be getting it...

I simply think you don´t get it

 

[AusWolf]

I simply think you don´t get it

Ok, let's be even simpler.

What a real or imagined "typical consumer" buys is completely irrelevant for a product review. What you're testing is the product itself, and not a specific hardware configuration. Therefore, you have to eliminate all other bottlenecks to let the product do its best and give its real performance.

Got it now?

 

[Lew Zealand]

Consumers which use an APU won´t buy biggest motherboards with the most power consuming chipsets

Idle power consumption is a big topic for me and I think it´s more realistic with an A-chipset-board/low entry B-chipset-board for AMD

Problem is, almost nobody cares or even thinks about chipset power use. Motherboard cost for matching low end parts? Yes, but not power use.

Otherwise, I could be interesting if Techpowerup also compare the power consumption differences of the tested mainboards in a graph for a quick overview

W1zzard has often said that if a thing doesn't get tested on TPU, it's because nobody's interested in testing it. You're gonna have to find someone to do this elsewhere.

 

[Vincero]

Problem is, almost nobody cares or even thinks about chipset power use. Motherboard cost for matching low end parts? Yes, but not power use.

That's not entirely true - I am hardly 'impressed' about AMDs AM5 "just use another IO/PCIe splitter chip" approach to segmentation as it is, even in very small terms, more wasteful in terms of power use let alone PCB space and probably adding to the BOM. Am I going to sulk about it? No, it's not like they added something with the power drain of a 2nd CPU, but they did cheap out a bit on board chipsets for this first generation of AM5 boards - even if they made a chip package with 2 IO/PCIe splitter dies together, that would have yielded a minor improvement.

In real terms the difference between say the worst AM4 chipset and the best is ~6W. Technically you could say the full 15W consumption of the X570 could be written off if you include the chipsetless motherboards that just use the CPU SoC IO only, but that's very niche so going to ignore it.

As I said before, a cheap motherboard using the leanest A/B series chipset but fitted with crappy VRMs and other board components could easily end up drawing more power than a good X570 board, assuming all other specifications are equal.

 

[Scour]

Maybe it´s country-depended but with the highest energy costs in Europe I think some ppl care about it.

 

[chrcoluk]

just wait for 2 years, and its gonna be super cool CPU, coz it will drop in price and motherboard too.

End of gen is best time to buy AMD, their prices do seem to consistently drop over time.

 

[rullydotcom]

I think this is the best CPU desktop in terms of performance/watt. But that x4 lanes tho ....

It would he nice to see the power consumption of the other two siblings (8700g & 8600g)