AMD Ryzen 5 2400G Review: Zen, Meet Vega

[reddragon55]

Soo, we're comparing the AMD Ryzen 2400G APU ($169) to the Intel i5-8400 ($189) + Nvidia GT 1030 ($120) = $310. Why? So Intel can have something at the top of the benchmarking charts?

Why not compare AMD's APU directly to Intel's APU? Ohh wait, the Intel one wouldn't be able to deliver even single digit FPS.. Shameful.

 

[PaulAlcorn]

I think it is not fair to include the cost of the graphics card in the intel ($ 89) and not include in Amd the extra cost of the memories to 3200 compared to the 2400/2666 (approximately $ 60).

AMD was only tested @ DDR4-2933 for stock (supported speed) and DDR4-3200 for the OC. None of the Intel chips can OC. We ran those at the supported DDR4-2400/2666.

Soo, we're comparing the AMD Ryzen 2400G APU ($169) to the Intel i5-8400 ($189) + Nvidia GT 1030 ($120) = $310. Why? So Intel can have something at the top of the benchmarking charts?

Why not compare AMD's APU directly to Intel's APU? Ohh wait, the Intel one wouldn't be able to deliver even single digit FPS.. Shameful.

Intel onboard graphics are in the 720p tests. From the review --

We're also testing the Core i5-8400's UHD Graphics 630 engine and AMD's Bristol Ridge-based A10-9700 at 1280x720. We didn't bother benchmarking them at 1920x1080, and you'll see why.

 

[TMTOWTSAC]

Hopefully, when we look back this could be the chip that raised the bar on minimum acceptable graphics. Intel has already decided their IGP wasn't enough and licensed AMD tech. The only reason cards like the 550 and 1030 were ever considered viable was that they outperformed IGP by such a large margin.

 

[network.learning.1000]

Where is the PC Version of Xbox one X APU AMD ?

 

[IInuyasha74]

Where is the PC Version of Xbox one X APU AMD ?

Well in terms of CPU power, this CPU is far superior to what's in the current game consoles. It just lacks the GPU power that the others have. We probably won't ever see that. At a certain point, it just doesn't make sense when you can always buy a GPU.

 

[InvalidError]

It just lacks the GPU power that the others have. We probably won't ever see that.

The 2400G is roughly comparable to a GT1030. If you go back ~10 years, that's a GTX280/285. Not many would have believed that this level of performance would become available as an IGP either.

IGPs are going to get really interesting when AMD and Intel decide to put HBM2/3 on-package: Iris Pro / GT3e was well ahead of the rest in the IGP department back then thanks to its 128MB 100GB/s L4 cache. Imagine what could be done today with 2GB of 250GB/s HBM2 (single 16Gb 1024bits-wide die) instead if AMD or Intel really wanted to to push integrated graphics. A ~$280 APU with an IGP approaching the 3GB GTX1060 should be feasible in the near future. Going down the CPU+GPU+HBM2/3 MCP route would also be an option to mitigate yield issues but may require a different socket to manage the different core voltages and extra power draw.

The technology is mostly already here, it is mainly a matter of what AMD and Intel will decide to do with it.

 

[chyaweth]

A question rather than a comment. Could I use the integrated graphics to drive three screens?

 

[mitch074]

A question rather than a comment. Could I use the integrated graphics to drive three screens?

Provided your motherboard has a DisplayPort output, you add a Displayport splitter and you allocate enough RAM to the Vega iGPU, then I'd tend to say yes.

 

[maddogfargo]

Honestly, I don't see non-metallic TIM as a con. Means they might be delidded. Fun. 😉

 

[VIVO-US]

That's some pretty good performance for integrated graphics, especially considering the price! There is one thing I want to know that I can't seem to find anywhere. How are these for multi-monitor support? Could they output to 3 monitors if the motherboard used has 3 video ports? If so, this would be extremely useful for office computers instead of having to use a separate GT 710 to get video on 3 displays.

 

[spdragoo]

I would have definitely liked to have seen some other Ryzen processors in these benchmarks, at least the 1400 and 1500X for this review, to see whether the architectural changes had any notable effects on performance.

Also, why not test the Ryzen 2400G with a GT 1030 as well, to get a more accurate indication of how the built-in graphics compare to that card?

Perhaps testing with 8GB of RAM might have been useful too, since RAM is rather expensive right now, and someone building a system with one of these for low-budget gaming probably isn't going to put close to $200 worth of RAM into it. With only 8GB, having some of that dedicated to the graphics system could potentially cause a hit to performance in certain games.

Not a whole lot of cost difference between DDR4-2400 & DDR4-3200 RAM -- only $2 USD difference for a G.Skills Ripjaws V 2x4GB set (https://pcpartpicker.com/products/memory/#s=402400,403000,403200&Z=8192002&sort=price). Nor is jumping up to 16GB that bad of a price -- looks like an extra $75 USD or so for the G.Skills (https://pcpartpicker.com/products/memory/#s=402400,403000,403200&Z=8192002&sort=price), so maybe 80% more cost for 100% more RAM.

Techspot did do some testing with dedicated GPUs for the APUs, however, & even compared them to the non-APU Ryzens (https://www.techspot.com/review/1574-amd-ryzen-5-2400g-and-ryzen-3-2200g/)...& not only did it show these APUs with a GT 1030/RX 550 matching or exceeding the performance of a Coffee Lake i3, they weren't very far behind their non-APU Ryzen 3/5 brethren.

 

[logainofhades]

Definitely a step in the right direction. With the jacked up pricing for a dedicated card, this gives the budget builder's foot in the door, allowing them some gaming, until the dedicated card pricing finally normalizes.

 

[antonio.parrot]

I think it is not fair to include the cost of the graphics card in the intel ($ 89) and not include in Amd the extra cost of the memories to 3200 compared to the 2400/2666 (approximately $ 60).

AMD was only tested @ DDR4-2933 for stock (supported speed) and DDR4-3200 for the OC. None of the Intel chips can OC. We ran those at the supported DDR4-2400/2666.

Soo, we're comparing the AMD Ryzen 2400G APU ($169) to the Intel i5-8400 ($189) + Nvidia GT 1030 ($120) = $310. Why? So Intel can have something at the top of the benchmarking charts?

Why not compare AMD's APU directly to Intel's APU? Ohh wait, the Intel one wouldn't be able to deliver even single digit FPS.. Shameful.

Intel onboard graphics are in the 720p tests. From the review --

We're also testing the Core i5-8400's UHD Graphics 630 engine and AMD's Bristol Ridge-based A10-9700 at 1280x720. We didn't bother benchmarking them at 1920x1080, and you'll see why.

First of all, congratulations for your work.
I am sorry. I have not explained well. The cost of a PC is the sum of the parts. The benefit is maximum when they are compensated. If I wanted to make a cheap PC with Intel to play in 1080 in low I would not take i5-8400 + Gt 1030 + 16 GB Ram but G4560 + GTX 1050 vanilla + 8 GB of Ram. Surely I would have more FPS.

I think nobody buys an i5-8400 to pair it with a 1030 gt.
Why have not they included the test of an R 1600 x + RX 550?

 

[cryoburner]

...Perhaps testing with 8GB of RAM might have been useful too, since RAM is rather expensive right now, and someone building a system with one of these for low-budget gaming probably isn't going to put close to $200 worth of RAM into it. With only 8GB, having some of that dedicated to the graphics system could potentially cause a hit to performance in certain games.

...Nor is jumping up to 16GB that bad of a price -- looks like an extra $75 USD or so for the G.Skills (https://pcpartpicker.com/products/memory/#s=402400,403000,403200&Z=8192002&sort=price), so maybe 80% more cost for 100% more RAM.

Sure, the price difference might not exactly be huge to move up to 16GB, but someone considering playing games on a GT 1030 / RX 550 level GPU will undoubtedly be cutting corners in the cost of their system wherever they can, and RAM currently costs around double what it did a little over a year ago, making it not a great time to buy more than you have to. And at least for the time being, with a dedicated graphics card, 8GB of RAM should be enough for pretty much any game to run reasonably well. Once you've dedicated a couple gigabytes of that to the integrated graphics though, you might potentially encounter performance issues in some games with just the remaining 6GB of system memory.

And while yes, they could spend more to move up to 16GB, that kind of cancels out much of the savings associated with going with the integrated graphics to begin with. And just as one could spend an extra $75 or so to move up to 16GB of RAM in a build, one could also spend less than that to stay with 8GB, but move up from a GT 1030 to a much more capable GTX 1050 instead, which would undoubtedly result in much better gaming performance than the extra RAM. As an example, for a roughly comparable cost to a Ryzen 2400G with 16GB of RAM, one could get a Ryzen 1200 or 2200G with 8GB of RAM and a GTX 1050, which would undoubtedly be better for gaming, at least in today's games.

Of course, it's possible that dropping accessible system memory to 6GB might not hurt performance that much in most titles, but it's difficult to say for sure without seeing that tested. At the very least, I think it's something worth testing in a review for these CPUs.

 

[nate1492]

First of all, congratulations for your work.
I am sorry. I have not explained well. The cost of a PC is the sum of the parts. The benefit is maximum when they are compensated. If I wanted to make a cheap PC with Intel to play in 1080 in low I would not take i5-8400 + Gt 1030 + 16 GB Ram but G4560 + GTX 1050 vanilla + 8 GB of Ram. Surely I would have more FPS.

I think nobody buys an i5-8400 to pair it with a 1030 gt.
Why have not they included the test of an R 1600 x + RX 550?

The answer is really, really, simple. There was a narrative that the author wanted to provide.

It falls so utterly flat, when the PC used has 16gb of very fast ram, a 144 dollar CPU cooler, and a 90 dollar motherboard.

The tests and excecution and comparison were designed to make the Raven Ridge look great against components no one would ever consider pairing together.

 

[cryoburner]

It falls so utterly flat, when the PC used has 16gb of very fast ram, a 144 dollar CPU cooler, and a 90 dollar motherboard.

If you were following along, the reviewer already pointed out that for the tests performed at stock clocks, they used the bundled cooler and ran the RAM at stock speeds. And you certainly don't need a $144 liquid cooler to get that kind of overclocked performance from these 65 watt CPUs. Even a $20 cooler like a 212 EVO would more than likely be plenty to overclock both the CPU and integrated GPU.

About the only thing I would agree on is that 16GB of RAM could potentially affect performance in some games compared to the 8GB someone would more likely have in a low-budget gaming system, when you take into account the amount set aside for the integrated graphics.

Also, if you want to get picky, why not point out that the Intel CPUs were all tested on $200 motherboards? Plus, you still can't even buy anything but enthusiast-level motherboards for Intel's 8000 series CPUs. Occassionally you might find one on sale for around $100, but for the most part, they're all $120 or more.

 

[InvalidError]

Why have not they included the test of an R 1600 x + RX 550?

The answer is really, really, simple. There was a narrative that the author wanted to provide.

There is no narrative: the motherboard, memory and CPU for the Raven Ridge review were provided by AMD so that's what it got "stock-benched" with just like every other reviewer out there who received AMD's reviewer kit.

 

[nate1492]

Why have not they included the test of an R 1600 x + RX 550?

The answer is really, really, simple. There was a narrative that the author wanted to provide.

There is no narrative: the motherboard, memory and CPU for the Raven Ridge review were provided by AMD so that's what it got "stock-benched" with just like every other reviewer out there who received AMD's reviewer kit.

The Intel comparison is where things go sideways, not the provided hardware.

Why compare a CPU with better ram against a CPU with worse ram? 2400 v 3200? That's 60+ dollar difference and massive performance changes. We've seen countless reviews showing just how much RAM benefits gaming (and Ryzen).

It really makes the comparison pointless and the only reason I can think of to compare the two is because using comparable hardware shows there is no benefit to Raven Ridge.

TPU did a great like for like and shows this. https://www.techpowerup.com/reviews/AMD/Ryzen_5_2400G_Vega_11/20.html

The 1030 v Raven Ridge results are identical even with 4 ghz OC. And they show what a 1050 does to it (hint, it smashes it). So, looking at these OC results, it's the RAM that actually changes the game performance, not the GPU or the CPU.

 

[rhysiam]

Why compare a CPU with better ram against a CPU with worse ram? 2400 v 3200? That's 60+ dollar difference and massive performance changes. We've seen countless reviews showing just how much RAM benefits gaming (and Ryzen).

It really makes the comparison pointless and the only reason I can think of to compare the two is because using comparable hardware shows there is no benefit to Raven Ridge.

I don't know where you're checking your prices, but as others have pointed out above, the price hike on fast RAM is nowhere near that much.

According to PCPartpicker at the time of writing: (excluding MIRs)
2x4GB Options
Cheapest 2400Mhz kit: $89: https://pcpartpicker.com/product/2gJkcf/gskill-memory-f42400c15d8gnt
There's a 3000Mhz Ripjaws V kit for $94: https://pcpartpicker.com/product/wTCrxr/gskill-memory-f43000c15d8gvrb
Or a 3200Mhz Ripjaws 4 Kit for for $103: https://pcpartpicker.com/product/njp323/gskill-memory-f43200c16d8grb
So that's $5-13 extra for the faster 2x4GB kit depending on your choice.

2x8GB Options
Cheapest 2400Mhz kit: $165: https://pcpartpicker.com/product/4vWrxr/gskill-memory-f42400c15d16gvr
There's a 3000Mhz Team Vulcan kit at the same price: https://pcpartpicker.com/product/jx8H99/team-vulcan-16gb-2-x-8gb-ddr4-3000-memory-tlred416g3000hc16cdc01
Or you pay $10 more for a Trident Z 3000 kit: https://pcpartpicker.com/product/ZskwrH/gskill-memory-f43000c15d16gtzb
Or $23 more for a Ripjaws 3200Mhz kit: https://pcpartpicker.com/product/qGqbt6/gskill-memory-f43200c16d16gvgb
So that's $0-23 extra for the faster 2x8GB kit depending on your choice.

There is a small cost difference between platforms here and that should absolutely be factored into the tight value comparisons that are being made at these sorts of budget. But it's nothing like $60.

Your other question is about "pointless" comparisons when Intel CPUs are running slower RAM. The reason some reviewers are putting 2400Mhz RAM on the Intel CPUs is because 8th Gen i3s are only listed to support memory up to 2400Mhz. Anything above that is technically overclocking which on Intel's platform requires a Z series motherboard. Once the B & H series mobos for 8th Gen are released, anyone using one of those motherboards with an i3 will be limited to 2400Mhz RAM (i5s & i7s go to 2666). With AMD on the other hand, now the early RAM compatibility issues are mostly resolved, a cheap B350 board is all you need for higher clocked RAM.

You could certainly argue that systems should be spec'd out identically. But others would counter that a more useful comparison is the "sensible" spec for each platform. In the long term, Z370 boards won't generally be the smart buy at these sort of price points, so the 2400Mhz RAM on Intel is a realistic comparison.

 

[King_V]

I don't know where you're checking your prices, but as others have pointed out above, the price hike on fast RAM is nowhere near that much.

According to PCPartpicker at the time of writing: (excluding MIRs)
2x4GB Options
Cheapest 2400Mhz kit: $89: https://pcpartpicker.com/product/2gJkcf/gskill-memory-f42400c15d8gnt
There's a 3000Mhz Ripjaws V kit for $94: https://pcpartpicker.com/product/wTCrxr/gskill-memory-f43000c15d8gvrb
Or a 3200Mhz Ripjaws 4 Kit for for $103: https://pcpartpicker.com/product/njp323/gskill-memory-f43200c16d8grb
So that's $5-13 extra for the faster 2x4GB kit depending on your choice.

2x8GB Options
Cheapest 2400Mhz kit: $165: https://pcpartpicker.com/product/4vWrxr/gskill-memory-f42400c15d16gvr
There's a 3000Mhz Team Vulcan kit at the same price: https://pcpartpicker.com/product/jx8H99/team-vulcan-16gb-2-x-8gb-ddr4-3000-memory-tlred416g3000hc16cdc01
Or you pay $10 more for a Trident Z 3000 kit: https://pcpartpicker.com/product/ZskwrH/gskill-memory-f43000c15d16gtzb
Or $23 more for a Ripjaws 3200Mhz kit: https://pcpartpicker.com/product/qGqbt6/gskill-memory-f43200c16d16gvgb
So that's $0-23 extra for the faster 2x8GB kit depending on your choice.

Yes, but acknowledging this fact would defeat nate1492's "narrative" of a $60+ RAM price difference.

 

[larkspur]

2x8GB Options
Cheapest 2400Mhz kit: $165: https://pcpartpicker.com/product/4vWrxr/gskill-memory-f42400c15d16gvr
There's a 3000Mhz Team Vulcan kit at the same price: https://pcpartpicker.com/product/jx8H99/team-vulcan-16gb-2-x-8gb-ddr4-3000-memory-tlred416g3000hc16cdc01
Or you pay $10 more for a Trident Z 3000 kit: https://pcpartpicker.com/product/ZskwrH/gskill-memory-f43000c15d16gtzb
Or $23 more for a Ripjaws 3200Mhz kit: https://pcpartpicker.com/product/qGqbt6/gskill-memory-f43200c16d16gvgb
So that's $0-23 extra for the faster 2x8GB kit depending on your choice.

There is a small cost difference between platforms here and that should absolutely be factored into the tight value comparisons that are being made at these sorts of budget. But it's nothing like $60.

I think most of the griping going on is because the FlareX kit used in this review represents a best-case scenario for these APUs. It's CAS 14 @3200. And it costs $250 on Newegg. As you linked, a DDR4 2400 kit CAS15 costs $165. That's a significant difference. The DDR4 3200 that you link for $188 is CAS16 which will not perform nearly as well as the FlareX kit (assuming it can even be run @3200 on a cheap Ryzen board). Despite the numerous firmware updates, many people are still having RAM compatibility issues especially at speeds over DDR4 3000. This is where the FlareX kits come in.

Of course, I don't blame the review at all since this is the review kit that AMD sent. It is however, reasonable to assume that very few people building with a Ryzen APU are going to spend $250 on RAM. Anyway, I just wanted to point out that I think that's what the griping is about. In other words, a builder won't get the level of OC graphics performance found in this review unless they overspend on RAM by ~$60. $60 is significant in entry-level builds.

The techspot review actually shows memory scaling on the Ryzen APUs very well and the differences appear to be significant. Not a deal-breaker, but significant nonetheless: https://www.techspot.com/review/1574-amd-ryzen-5-2400g-and-ryzen-3-2200g/page8.html

I really like the 2200g @ $99. That is a clear value winner and a great entry-level gaming recommendation - Great job AMD! But I don't see enough performance delta with the 2400g to justify the $70 price difference for the entry-level segment.

 

[logainofhades]

I really like the 2200g @ $99. That is a clear value winner and a great entry-level gaming recommendation - Great job AMD! But I don't see enough performance delta with the 2400g to justify the $70 price difference for the entry-level segment.

I have to agree. The 2200g also makes the 1200 and 1300x irrelevant now. As nice of a chip as the Pentium G4560 is, the 2200g makes far better sense. It can overclock, it's a quad core, has better graphics, and is on a current gen platform.

 

[King_V]

The techspot review actually shows memory scaling on the Ryzen APUs very well and the differences appear to be significant. Not a deal-breaker, but significant nonetheless: https://www.techspot.com/review/1574-amd-ryzen-5-2400g-and-ryzen-3-2200g/page8.html

That's definitely intriguing information. Though, I do have to object to this because, from everything I've read, CS:GO is heavily CPU dependent. Further, they ran it at 720p. Would the differences in performance for the various memory speeds be be as pronounced, percentage-wise, at 1080p? How about with a game that's more CPU dependent than GPU?

You also made mention of the difference of the CAS14 vs CAS16 for the FlareX vs Ripjaws V... but, in gaming usage, would that have anywhere near as pronounced an effect as what was demonstrated on the various RAM speeds? More? Less? I don't know, but am genuinely curious.

 

[InvalidError]

That's definitely intriguing information. Though, I do have to object to this because, from everything I've read, CS:GO is heavily CPU dependent.

Any game will look "CPU-heavy" if you run a lightweight game with no frame rate limit where the GPU/IGP is a non-issue. If CSGO was actually CPU-intensive, you wouldn't be seeing 110+fps average and 80+fps 1% lows on an entry-level CPU.

 

[rhysiam]

I think most of the griping going on is because the FlareX kit used in this review represents a best-case scenario for these APUs. It's CAS 14 @3200. And it costs $250 on Newegg. As you linked, a DDR4 2400 kit CAS15 costs $165. That's a significant difference. The DDR4 3200 that you link for $188 is CAS16 which will not perform nearly as well as the FlareX kit (assuming it can even be run @3200 on a cheap Ryzen board). Despite the numerous firmware updates, many people are still having RAM compatibility issues especially at speeds over DDR4 3000. This is where the FlareX kits come in.

Expensive low latency RAM has been shown to help with certain CPU tasks, but gaming performance on these APUs will almost always be GPU limited. Graphics memory is all about bandwidth, not latency. GDDR5 is designed with this very use-case in mind, trading off low latency for a significant increase in bandwidth. Even a 2133Mhz CL14 kit will have lower latency than most (all?) GDDR5 implementations. If anyone has seen trustworthy benchmarks looking at graphics scaling with RAM timings I'd happily take a look, but I'd be very surprised if the latency makes any tangible difference for gaming.
Memory bandwidth/throughput... absolutely important! Just not latency. So those cheap 3000/3200Mhz kits are, I suggest, the right price/performance pick for one of these APUs.

From what I understand the vast majority of memory compatibility issues are resolved on Ryzen now (at or under 3200Mhz), particularly with common kits (which is why I linked Ripjaws & Trident Z options at each capacity point). YMMV I guess.