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

[dungeon]

The 2400G consumes around 50 W in real-world gaming according to Tomshardware. This means about 30% more headroom to stay inside 65 W TDP, and power consumption can be almost doubled and still stay inside the 95 W limit of the AM4 socket.

Realworld gaming means nothing as usually that is not scenario where max spikes are, they need to more stress - something like running Prime + some GPU benchmark (Unigine Valley let say) at the same time.

That is what i recommend running to test OC stability even and for two hours minimum

Since this is an APU and in addition to Prime + Valley it won't hurt if they decode some movie at the same time also (or whatever else to fire up all blocks there), just to be more right at theoretical spikes of an APU

3/4 of Prime + else (leave it one core for else), as else wanna use some CPU too... might be an ideal stress scenario combo for APUs. Just watch some movie for 2 hours and let it do Prime + Valley behind the scene and if nothing smells or get stuck - everything is fine

Of course make sure you use hardware decoding to play that movie as point of stress test is to put miltimedia block on fire too

[sykobee]

Looks like you'll be getting $20 back from NewEgg for the 2400G overprice - https://www.techpowerup.com/241525/n...ding-customers

[konserw]

Michael - could you please share link to openbenchamrking ? I'd like to compare my system with those

[Michael]

Michael - could you please share link to openbenchamrking ? I'd like to compare my system with those

You can get the ID from looking at the graph links but it's 1802130-FO-GRAPHICSR44

[artivision]

Someone should try to unlock Vega-8 into 11-12.

[Alliancemd]

Eh, as this is a real APU, and as long as the driver support is good enough : this might make very little difference? if you run out of "graphics memory" you can read textures from main memory (since the AGP 2x days), and here both the CPU and GPU have about the same way of accessing memory, through that big coherent interconnect.

So if you have no weird overhead, no access violation, I think you'd see identical results whether the setting says 512MB or 2GB. Unless you would get an out of memory condition with 512MB sometimes (three 4K displays, triple buffered, FP16 framebuffers, buffers you're rendering into...) I think the setting would mostly tell the OS this area is "reserved" so that it doesn't put programs and disk cache in there?
/edit : or it will be slower depending on number of memory allocations..

Ok, finally found where I saw that: https://youtu.be/H1FUKgOh1pM?t=12m

[grok]

I'll be waiting for something native, not wanting to use nVidia or a discreet card, Raven Ridge sounds perfect for my needs, just want it in the smallest form factor possible

Some B450 chipset motherboard maybe? Surely many people will want that feature so it really should be found on the new boards.
My point otherwise is a DP to HDMI dongle would be no less native than same chip on motherboard, but it sticks out outside and costs $$. (you end up with two hdmi outs, one 1.4 and one 2.0)

I like to be overly exhaustive. say you have some really great rig like i5 2500k or better and 24 gigs ddr3, a GT1030 would be good if hdmi 2.0 is needed (depending on middle finger to nvidia)

[FireBurn]

It'll mostly be for playing video and some light web stuff, a Ryzen APU would be perfect. Plus as you say, it gives the middle finger to nVidia. I've been Intel or AMD CPU and AMD only graphics for over a decade now

[yurikoles]

Michael there is a typo:

The Kaveri API

[yurikoles]

Michael you had compared game speed with CPU speed:

The Talos Principle was also much faster than the Intel graphics