DDR4 Memory Scaling Performance On AMD Raven Ridge / Ryzen 5 2400G

[dungeon]

I think Michael should currently test stock speeds only, as if stock does not look fine what one can expect from any overclocking - that would be even worse.

One need first to make sure utter stable driver is on stock speeds, as that is predependency to even start to play pushing around

[stiiixy]

I would buy faster RAM for APU if it yielded faster GFX performance. To a point based on cost/benefit at the time. You would be wasting the APU if you didnt, but the price of RAM at the high end might as well be spn t on a low end system using discrete. Pity we cant have the benchies yet and I look forward to seeing some. Certainly interesting times at AMD.

[sdack]

I think Michael should currently test stock speeds only ...

I think Michael has a mind of his own and only because of it do we have Phoronix. So I think he should just keep doing what he does.

Now me personally, I don't want to see what a CPU is capable of only today, but I want to see what it is capable of within the foreseeable future. It's still an investment to me and not an instant gratification. I do like to know what I can expect from a CPU in the near future. I've also seen how the Ryzen 5 2400G compares to the FX8350 and I bet the FX8350 was running with some decent DDR3 memory modules. Another good reason why Michael shouldn't limit his tests, but keep testing it on the cutting edge as well.

[dungeon]

I think Michael has a mind of his own and only because of it do we have Phoronix. So I think he should just keep doing what he does.

Now me personally, I don't want to see what a CPU is capable of only today, but I want to see what it is capable of within the foreseeable future. It's still an investment to me and not an instant gratification. I do like to know what I can expect from a CPU in the near future. I've also seen how the Ryzen 5 2400G compares to the FX8350 and I bet the FX8350 was running with some decent DDR3 memory modules. Another good reason why Michael shouldn't limit his tests, but keep testing it on the cutting edge as well.

Micheal could keep informing us on current situation i value that, but if it is situation that even stock does not work stable what you expect you will get from overclocking - that won't fix issues just potentionaly to raise them also you can't conclude anything for foreseeable future by looking at weird results

In foreseeable future we might get another Meltdown/Spectre alike vulnerabilities, so we again would need some software mitigations... it is same for games, one version might perform like this today but tommorow after some update might perform who could guess how... basically you can't conclude anything about future when it comes to performance, that might be always better but also worse

These APUs are pretty much all unlocked, one might want to OC it, another might don't want that but to put it down on low power even and third would just run it on stock - so there are pretty much vague areas of performance interests there There again, if stock is not stable what you expect from it?

One can run tests on stock and add plus/minus 40% potentional and theoretical push up and push down extremes, so that is what it is is capable of (so there is a possibility, but there is no warranty on anything there when it comes to OC, so you can't conclude anything again). They ship it at optimal speed and now you can play with it of course since it is unlocked.

[sdack]

Micheal could keep informing us on current situation i value that, but if it is situation that even stock does not work stable what you expect you will get from overclocking - that won't fix issues just potentionaly to raise them also you can't conclude anything for foreseeable future by looking at weird results

In foreseeable future we might get another Meltdown/Spectre alike vulnerabilities, so we again would need some software mitigations... it is same for games, one version might perform like this today but tommorow after some update might perform who could guess how... basically you can't conclude anything about future when it comes to performance, that might be always better but also worse

These APUs are pretty much all unlocked, one might want to OC it, another might don't want that but to put it down on low power even and third would just run it on stock - so there are pretty much vague areas of performance interests there There again, if stock is not stable what you expect from it?

One can run tests on stock and add plus/minus 40% potentional and theoretical push up and push down extremes, so that is what it is is capable of (so there is a possibility, but there is no warranty on anything there when it comes to OC, so you can't conclude anything again). They ship it at optimal speed and now you can play with it of course since it is unlocked.

The test results don't look that wrong to me. I was expecting them to scale and they do, so that's nice to know.

Not sure why you mention Meltdown/Spectre. Is it a bait argument? It's not like such issues are found everyday - it took a group of researchers to figure it out - and it's a problem for a lot of CPUs, not just this one. And I do remember AMD saying that theirs don't suffer as much from it than others.

Anyhow, it's not hard to see the point Phoronix is making here. Only yours is.

[dungeon]

It is an argument that Raven APUs are currently vague or fifty-fifty unstable even on Windows, so people still invesigate ways how to make to not be so unstable

[starshipeleven]

It is an argument that Raven APUs are currently vague or fifty-fifty unstable even on Windows, so people still invesigate ways how to make to not be so unstable

Damn windows, it needs ABi stability to be competitive with Linux.

[linuxgeex]

It is an argument that Raven APUs are currently vague or fifty-fifty unstable even on Windows, so people still invesigate ways how to make to not be so unstable

Thanks for that. The fellow doing the presentation doesn't do himself any favours with his non-specific and dismissive attitudes throughout, but if you can ignore that lack of professionalism on his part he actually has some helpful information in the last 4 minutes.

[Ansla]

Wasn't one of the improvements of DDR4 that multichannel configurations would be easier to implement than for DDR3? Because triple or even quad channel should really benefit these APUs.