Ryzen CPU + Vega Graphics on a Chip: AMD Ryzen 5 2400G & Ryzen 3 2200G Review

LOL, you should seriously work for AMD marketing. Sorry but 'cutting edge and vibrant' and... BANDWITH LIMITED. You can evolve drivers 100 times but nothing will get past the hard limit of simple memory bandwith, which will forever render these APUs to 720P gaming unless 1080P low settings at 30fps is your kind of 'cutting edge' gaming ;)

If given the choice, I'd take even an old dGPU over an APU anyday, and I don't care if the older dGPU is 'amortised, not cutting edge, non vibrant and isn't evolving rapidly'... haha

You actually quote me, then immediately claim I said something entirely different. That's sad.
 
How about hybrid crossfire on ryzen gpu vs discrete gpu (rx550/560) in dx12 (like Wolfenstein II - why you don't test it o_O)?
 
Here’s a question. What does all this “sync” stuff do outside of games? Such as making DVDs or BDs better? Or Word? Or SQL?
I’m the first to say games drive innovation in tech, but what does all this sync stuff do for anything beyond games.
Btw, it’s a real question; not sarcasm or anything. I’m trying to see the ‘other side’ so to speak. What will this do for us in the future?
Games drove memory leading to crystal clear 1080p They drove HD and frame rates giving us UHD 60, SHD-120, etc. I’m lost on what sync will do for non games.
 
The APUs should be enough to not bottleneck a GTX 1060 (or comparable AMD GPUs) in most games, and after that you may experience varying degrees of bottlenecking (depending on the game etc.) when moving to better GPUs. I don't think you'd hit a 100% CPU bottleneck with any GPU, but considering how expensive something like a GTX 1080 or a GTX 1080 Ti is, I'd find it hard to justify getting one to myself without upgrading the CPU as well. A GTX 1070 would still give you noticeable gains over a GTX 1060, but it may be that you wouldn't be able to find one at a reasonable price.



Even with a GTX 1080 the difference between x8 and x16 is in practice margin of error stuff and totally imperceptible, so there's absolutely no reason to expect the RX 550 would be crippled by x8.



GDDR5 has a higher effective clockspeed and larger bus width, which together mean a larger memory bandwidth. The memory bandwidth of the iGPU depends on the system specs, but we're probably talking about roughly an order of magnitude difference. Comparing the effect of DDR4 and GDDR5 speeds is complicated by the fact that with the APUs, the DDR4 speed also affects the performance of the CPU, not just the iGPU.



I don't think that's the correct question. A better one would be: Who would buy an RX 550 if they already have one of these APUs? In my opinion the minor performance gain is not really worth it. Either go for an RX 560 or a used GPU for the price of an RX 550.

If I were to lean toward a beefier dgpu system after agonising over an apu, I would second the choice of the ~$165 4GB rx560, along with a 1600 (or 1600x preferably), and a 400 (new series) am4 mobo.

The 560 is Polaris, a generation ahead of the 550, so the 560 is a closer sibling than the 550, and so is closer to the appealing apu ecosystem.

So you spend ~$340 on processors vs 170$ for an apu, but can save ~$80 on ram as you can better get away with 8GB of ddr4 w/ the dgpu. For ~$100 more, you get significantly more cpu & gpu cores, & 4GB of fast (100GB/s?) gpu cache.

Boys being boys tho, we like to think we are buying a tool rather than having fun, and making the apu sing (which it can easily), sure sounds fun.

AMD's utterly modern zen and Vega integrated on one easily cooled and unobtrusive die, is almost identical to the most advanced amd g/cpu AI mega server, except in scale, so there is much to explore & tools to do it with. A 560 Polaris dgpu places you a GPU generation behind.

A seismic difference is that by consolidating most of the important resources onto a ~tiny apu die, they have reduced their playing field from a soccer pitch to a pingpong table size, and all the old restrictions on the speed the ball can be passed around change. AMD can have their own, smarter & faster alternative rules in Infinity Fabric, or still use the old pci rules for now if it suits.

AFAIK the apu'S Fabric bus has; cpu, gpu, mem controller, io, ..., and more and then more again on future apuS - all in the footprint of a ~credit card. There will be many former restrictions that coders can bypass on an apu.
 
Just saying, but for AM4 fence sitters considering a $100 garbage rx 550 or Nvidia 1030 place warmer dgpu for their 4 core amd 2400g CPU, an option I have not seen mentioned, is to get a cheaper $100 apu til better apuS arrive. ie., get a placewarmer apu rather than a gpu.

AFAIK, amd still have plenty more TDP envelope to play with (AM4; 2400g=65w tdp, 1800x=95w tdp) . They pack a lot in, but they are easily cooled. Surely more powerful apuS are coming.

This makes strategic sense for amd, as the priority apu segment to address was mobile, and the modest initial 2200 & 2400G desktop apuS, are just upclocked variants of those same minimalist mobile apuS.

Preparing a good foundation the future upmarket apu, may superficially seem like overspending given the cheap main component.
 
I'd like to see how the CPU temperature scales when memory speeds increase. Starting @2133MHz all the way up to 3200MHz on the 2200/2400G CPU's.

I've been able to get my 16GB kit to run @ 2934MHz but Windows did reboot a couple of times. The CPU temperature goes up about 20C degrees between 2133MHz and 2934MHz.

I'm not sure if that's normal or not. Will really bad memory timings cause unnecessary temperature increases?
 
Back