Ryzen 5 2400G Radeon Vega Linux OpenGL/Vulkan Gaming Benchmarks

[cbxbiker61]

Amd sells Intel Gpus with 1300-1500 cores wile them selves still use 500-700, what a disgrace.

The Intel/AMD dual chip solution is not going to be cheap....rest assured that AMD will make money on each one. BTW, there is more space available on that package, so it's not surprising they have significantly more cores.

[dungeon]

Amd sells Intel Gpus with 1300-1500 cores wile them selves still use 500-700, what a disgrace.

That Intel + Vega is mobile Vega with HBM, separate silicion really... but this is just traditional APU.

These 24 CUs alone on mobile Vega (so even if you forget additional HBM size) is impossible to be in an Ryzen APU... but as you see 11 CUs like on this one could be packed.

On Playstation 4 PRO APU, much bigger gfx could be packed thanks to the small size of Cat CPU architecture so Jaguar CPUs size and also that GDDR5 memory is elsewhere

No memory makes things cheap, GDDR5 still makes things cheaper, but is big and eats power. HBM is expensive but product could be fast but smaller. But not a way to be in traditional APU design, particulary if you need to fit into classic AM4.

Again me with pictures with that Intel+VegaM(+HBM), how you will pack 24 CUs from the left into CPU on the right, as it is is obviosly bigger?

[wizard69]

Nice work Micheal!

It is just too bad a stable distro is so far off.

[khnazile]

Nice, but where are kill-ryzen.sh test results?

[faldzip]

This article is missing Pentium G4560 + Nvidia GT 1030 as this would be a same price range competitor to R5 2400G. Of course, when we are talking about gaming, not pure CPU perf, but that's what was tested here.

[Cape]

Oouw... Not the results I was expecting... I saw windows bench were the 2400G was almost 3 times faster than Intel!
Damn drivers..

[ElectricPrism]

Oouw... Not the results I was expecting... I saw windows bench were the 2400G was almost 3 times faster than Intel!
Damn drivers..

If it is a driver issue then we could expect correction, I'm sure more will be revealed in the days to follow when Michael does more tests on it.

When RX 480 came out performance went from 100% to 125% in 1 year, so I am curious if similar room for improvement or more is there and will get shaped up this year.

[ElectricPrism]

I'm a bit surprised the Iris Pro still holds up so well. Regardless, these are pretty good results for such a young GPU on Linux.


Actually depending on the chassis you get, ITX can often yield better temperature management. In a case like mine, the GPU fans get to breathe directly from outside the case with very little obstruction, while the CPU gets it's own wind tunnel from the front of the case and straight out the back. Might not be the most space efficient ITX design, but it is great at controlling temperatures without getting too loud.

I prefer thin cases that are as minimal as possible.

My first mod was this one:



In this case there were 2 SSD's under the motherboard mounted flat.

I have since removed both 3.5" HDD's, both SSD's and have switched to M.2

I have also switched from 3.5" to 2.5" SFX power supply



I am working on a new case about the same height and width of a a Xbox One S except with a slightly longer depth with a full dGPU. 2.5".

I basically have 3 options -- I can go buy a PicoPSU and do a AMD Ryzen 2400g build say with a NFC S4 Mini -- do a Dr Zaber Sentry although the costs is $360 to send to the states and I'm not 100% sure the dimensions are as small as they could be since my case is about 7.1 Liters of volume and IIRC it is a bit more than that.

[Alliancemd]

Reviewers say that it comes with a small 512MB VRAM preset in the BIOS.
Increasing it to 2GB increases the performance.
Did you run it with the default?

[_Alex_]

But there is funny fact there, as you can see they have one driver suite packaged and only for Windows 10 Now imagine we have only one Linux and one version of Linux to support, that one and only Linux is also up to date,etc... they will package that right away - but on Linux there is 300 distros and probably 300K different combinations In our world, one bug easely could be seen as 100 bugs, etc...

They could at least send some patches for the kernel, so that 4.15 can run this, instead of 4.16 / 4.17... I mean the hardware is ready for months for internal testing, it's not like the drivers need half a year or more.