[Hardware Unboxed] Ryzen 2400G/2200G Vega Graphics Performance

[tamz_msc]

Something is obviously wrong with the SC2 results in the pclab.pl results - it takes someone to be clinically blind to not notice that the 5675C with the Iris Pro doesn't even run the benchmark.

Like I've said before, taking into account the most popular games in Steam as well as BF1(Origin) and Overwatch(Blizzard) the 2400G beats or ties the G4560+GT 1030 in more games than it loses.

 

[IllogicalGlory]

There may be some truth to what 24601 is saying about Ryzen's performance in Starcraft 2, even if he's completely insufferable about it.

https://www.youtube.com/watch?v=fOrPfXGvugs

Ryzen 7 1700 3.8 GHz: Average framerate 27.8 FPS, Minimum framerate 10.9 FPS
Core i3-8350K 5.0 GHz: Average framerate 53.3 FPS, Minimum framerate 22.9 FPS

This video shows a Ryzen 1700 at 3.8GHz vs an i3-8350K at 5GHz, and in these really heavy scenes, the Ryzen side will drop down to below 20 FPS, while the 8350K maintains about 40-60. The game uses (almost) literally one thread, and the game behaves badly anyway.

PCGamesHardware.de Starcraft 720p tests:

http://www.pcgameshardware.de/Ryzen-7-1800X-CPU-265804/Tests/Test-Review-1222033/

 

[Red Hawk]

I have spoonfed you 11 times in a row now.

Do you even know what a "discussion" or "debate" is?

Why do you keep stating rhetorical statements disguised as questions?

To what degree do you lack the ability to present an actual argument instead of begging people to spoonfeed you?

Can't answer?

Guess you are simply wasting my time then?

Why do you keep getting offended when I spoonfeed you the things you clearly do not know anything about?

Ok. Let's back this up.

You're saying in absolute worst case scenarios in Starcraft 2, where the user is specifically trying to fill the screen with as many units and particle effects as possible, the 2400G's CPU will tank below 30 FPS. That's possible. But we have all of one benchmark indicating that, and you can't seem to be bothered to find more to corroborate those results. Then you do things like refer to Skylake's IMC in comparison to Ryzen's, that Ryzen's is licensed, which...I don't care about. Whatever is going on under the hood, I'm talking about the practical results here. With regards to that, in virtually every modern game, Ryzen is not bottlenecked to the point that it drops below 30 FPS. That calls the Starcraft 2 results into question.

Even taken at face value, what does that say about Ryzen? That if you artificially push a nearly 10 year old DX9 game on it as far as it will go, it will choke? Ok, sure. I don't know how useful that information really is, though. Much more relevant in 2018 is how the CPU does in modern DX11/DX12 titles that incur a considerable CPU load just through normal use.

 

[frozentundra123456]

Ok. Let's back this up.

You're saying in absolute worst case scenarios in Starcraft 2, where the user is specifically trying to fill the screen with as many units and particle effects as possible, the 2400G's CPU will tank below 30 FPS. That's possible. But we have all of one benchmark indicating that, and you can't seem to be bothered to find more to corroborate those results. Then you do things like refer to Skylake's IMC in comparison to Ryzen's, that Ryzen's is licensed, which...I don't care about. Whatever is going on under the hood, I'm talking about the practical results here. With regards to that, in virtually every modern game, Ryzen is not bottlenecked to the point that it drops below 30 FPS. That calls the Starcraft 2 results into question.

Even taken at face value, what does that say about Ryzen? That if you artificially push a nearly 10 year old DX9 game on it as far as it will go, it will choke? Ok, sure. I don't know how useful that information really is, though. Much more relevant in 2018 is how the CPU does in modern DX11/DX12 titles that incur a considerable CPU load just through normal use.

"Relevant" depends on what games you play. I play starcraft and other blizzard games, but play few if any DX 12 games, so it is relevant to me.

 

[frozentundra123456]

Did you even checked reviews from other sites? If yes, you would agree with me, and not post this.

I seriously doubt that. And I will post whatever I wish that meets forum rules, thank you. What I said is perfectly true. If there is anything false in that post, please point it out.

 

[Muhammed]

You forgot this one:

One game vs half a dozen games. We know Wolf2 works better on AMD hardware, the difference you are seeing there is down to vram capacity differences, as wolf 2 loves vram. the 1030 has 2GB, while Vega 11 has 4GB.

The rest of the games: GTA 5, StarCraft 2, OverWatch, World Of Tanks, Doom , Dirt Rally and others have no preference for VRAM, and yet they show a huge difference in these important titles in favor of the 1030, a difference that forces any serious gamer to consider the 1030 over Vega 11.

Results have to be read, understood and analyzed.

Numbers are valid across multiple sites, Vega 11 sucks in several high profile games, maybe try to analyze that instead of trying hard in a true fanboy fashion to skirt around the hard numbers.

 

[tamz_msc]

So let me repeat this again, taking into account those games which were tested from the top 20 on Steam, the 2400G performs in the following manner compared to the GT 1030:

• PUBG - loses but playable(30fps vs 40fps in 99%tile min)
  
• Dota 2 - loses but playable(same fps as above)
• CS:GO - tie
• R6:Siege - win
• GTA V - essentially a tie in terms of playability
• Civ 6 - tie
• Rocket League - tie according to AT, loss according to Hardware Unboxed.

Adding two more very popular titles -

• Overwatch - tie
• Fortnite - a very small loss

So out of the nine games I've mentioned here, all of them being highly popular titles right now, the GT 1030 is definitively faster than the 2400G Vega 11 iGPU in just two of them.

Yet there are some who claim that the GT 1030 is unequivocally better than the 2400G.

 

[Blitzvogel]

Now imagine if the Vega 11 and 8 were not hampered by limited bandwidth. The 2400G's graphics is on par with a base model PS4 at full graphics boost.

Too many people are touting these APUs as a work around with the current GPU pricing climate, but they're too bandwidth limited. Unless AMD's contract with Intel for Kabylake-G says otherwise, AMD should also get on the single package CPU-GPU-HBM game or produce a higher end APU with stacked HBM. It would help get the lower end gaming market get around the GPU supply problem.

 

[Muhammed]

GTA V - essentially a tie in terms of playability

Testing at very high settings is counter productive, both score sub 30. Test at medium and this is the result:

PUBG - loses but playable(30fps vs 40fps in 99%tile min)

A very important win. A 30% difference in favor of the 1030 in this popular title.

Overwatch - tie

Yet on another map, the 1030 dominates:
https://www.purepc.pl/procesory/tes...ven_ridge_zen_i_vega_w_jednym_ciele?page=0,15

So out of the nine games I've mentioned here, all of them being highly popular titles right now, the GT 1030 is definitively faster than the 2400G Vega 11 iGPU in just two of them.

There are other games you didn't list: Doom, World Of Tanks, Dirt Rally, StarCraft 2 and others.

 

[Blitzvogel]

Testing at very high settings is counter productive, both score sub 30. Test at medium and this is the result:

There are other games you didn't list: Doom, World Of Tanks, Dirt Rally, StarCraft 2 and others.

I've been noticing this issue with Raven Ridge scores. Hard to get a real picture without testing a multitude of configurations and resolutions for each game.

 

[Red Hawk]

Now imagine if the Vega 11 and 8 were not hampered by limited bandwidth. The 2400G's graphics is on par with a base model PS4 at full graphics boost.

Too many people are touting these APUs as a work around with the current GPU pricing climate, but they're too bandwidth limited. Unless AMD's contract with Intel for Kabylake-G says otherwise, AMD should also get on the single package CPU-GPU-HBM game or produce a higher end APU with stacked HBM. It would help get the lower end gaming market get around the GPU supply problem.

Unfortunately, this is pretty true. Just compare the RX 550 with the Vega 11 Graphics in the 2400G. The RX 550 has 8 Polaris CUs/512 stream processors, vs the 11 Vega CUs/704 stream processors of the Vega 11. Yet the Vega 11 doesn't hold a total advantage over the 550, even with fast DDR4. Raven Ridge appears to be just as bandwidth constrained as Llano was 7 years ago. Even though system RAM has gotten faster, the memory demands of modern games have advanced as well.

The solution may indeed be HBM memory. But that would make for an expensive, niche chip. I don't know if it's worth it to AMD to develop that themselves.

 

[Mr Evil]

I'm somewhat interested in these for use in a small server that needs a video output. Since they use the AM4 socket, it seems likely that they should support ECC RAM with the right motherboard, like the existing Ryzen CPUs.

 

[whm1974]

Personally I think that 1080@~45fps set to medium graphics is playable for the Ryzen 5 2400G. While it doesn't completely replace a decent dGPU, it is something to by on until prices returned to reasonable levels.

 

[yepp]

One game vs half a dozen games. We know Wolf2 works better on AMD hardware, the difference you are seeing there is down to vram capacity differences, as wolf 2 loves vram. the 1030 has 2GB, while Vega 11 has 4GB.

2GB vram is more than enough for Wolfenstein II at low 720p. It's more of a case AMD graphics just performs better in this game, same with Hitman.

https://www.youtube.com/watch?v=q1B_jmk6B9U
https://www.youtube.com/watch?v=DXK0hElsrzo

 

[tamz_msc]

A very important win. A 30% difference in favor of the 1030 in this popular title.

Okay NVIDIA can take that one.

Testing at very high settings is counter productive, both score sub 30. Test at medium and this is the result:

That test shows that both can achieve a locked 30fps by dropping a few settings. Basically a tie.

Yet on another map, the 1030 dominates:
https://www.purepc.pl/procesory/tes...ven_ridge_zen_i_vega_w_jednym_ciele?page=0,15

A bot match compared with the training arena? Really?

There are other games you didn't list: Doom, World Of Tanks, Dirt Rally, StarCraft 2 and others.

Okay so let accept NVIDIA+Intel winning in single-threaded SC2. Doom and Dirt Rally aren't in top 20. World of Tanks Encore is a benchmark which isn't representative of actual gameplay, so that doesn't count.

So that's two - SC2 and PUBG. Witcher 3 is hovering just outside the top 20 and pclab.pl also shows those two are indistinguishable in that game. Let's add that one as well.

So, GT 1030 wins conclusively in 3 out of 11 of all the above games we have considered. Still doesn't make it a better sale over the 2400G.

 

[Guru]

I'm talking about the AMD APU vs intel CPU + 550

UHD 630 is obviously GPU bottlenecked.

2400/2200g is just worse ryzen with less cores and less clocks (as well as getting reamed by memory bandwidth due to having to share with the iGPU)

I'm going to assume that you are just silly, otherwise that is absurd beyond belief or trolling.

The Ryzen 2400G is faster than Ryzen 1400, heck its actually faster than the Ryzen 1500x. Its got faster frequency starting at 3.6GHz and boosting to 3.9GHz, much more than the 1500x and the 1500x is faster than the I5 7500.

 

[Paratus]

The 2200G and 2400G are absolutely great products.

Intel can match the CPU performance but is completely outclassed in GPU performance.

Intel + NV can’t match the APUs combination of:

• CPU performance
• GPU performance
• Price
• Power usage.

For the Intel + 1030GT combo you have to give up 2-3 of those to win the other 1-2.

Despite what some folks keep shoveling (spooning? ) around here these APUs are very attractive for low cost boxes that can still play some games while being productive for a bargain price, even when including the price difference between DDR 3200 and DDR 2400.

I’d expect these in the short term to pick up a small amount of market share from Intel as they cannot compete (yet) with Vega integrated graphics and from NV in the low end discreet GPU market as the performance is good enough and NV cannot compete with AMD in the APU space.

 

[Muhammed]

That test shows that both can achieve a locked 30fps by dropping a few settings. Basically a tie.

This isn't a race to 30fps, GPU benchmarks are won by higher fps. Not lower!

 

[SirDinadan]

Raven Ridge appears to be just as bandwidth constrained as Llano was 7 years ago.

I've just finished a benchmark session with multiple games where I compared the average FPS result with the same settings but with different resolutions - 720p | 900p | 1080p. When upping the resolution Vega 8 shows similar performance drop compared with the GT 1030 and RX 550. I'm no expert, so correct me if I'm wrong - memory bandwidth is less important in lower resolutions as long as the GPU cores are fully utilized in all tested resolutions.

 

[tamz_msc]

This isn't a race to 30fps, GPU benchmarks are won by higher fps. Not lower!

When the differences are so small when using the highest settings, it means both of them provide an identical end-user experience. The GT 1030 typically winning by <10% doesn't matter in the end at this performance level(30fps)

 

[Qwertilot]

The 2200G and 2400G are absolutely great products.

Intel can match the CPU performance but is completely outclassed in GPU performance.

Intel + NV can’t match the APUs combination of:

• CPU performance
• GPU performance
• Price
• Power usage.

For the Intel + 1030GT combo you have to give up 2-3 of those to win the other 1-2.
.

Think it's mostly cost they have on their side actually. Which is nice enough

Power usage has gone a bit wrong somewhere. See AT review. Mobile Ryzen worked out quite nicely so not quite sure why.

When gaming it hits over 100w in the guru3d review. You can fit an awfully fast Intel cpu + 1030 into that budget. 1050 even maybe.

No big surprise as Ryzen is ~= to Intel power wise bit but vega really isn't vs Pascal (let alone the next gen.).

 

[Tapoer]

Think it's mostly cost they have on their side actually. Which is nice enough

Power usage has gone a bit wrong somewhere. See AT review. Mobile Ryzen worked out quite nicely so not quite sure why.

When gaming it hits over 100w in the guru3d review. You can fit an awfully fast Intel cpu + 1030 into that budget. 1050 even maybe.

No big surprise as Ryzen is ~= to Intel power wise bit but vega really isn't vs Pascal (let alone the next gen.).

On Guru3D review they used a 1200W PSU and the "over 100W figure when gaming" is for the whole system.

In the tomshardware review the 2400 used on average ~40-45W when gaming in The Witcher 3:

.

And ~50W on Final Fantasy benchmark:

But it seems that the TDP limits are not activated or working since it reached over 100W when using Prime95 + kombustor.

 

[Qwertilot]

I don't think the power limits are quite working if it's only running at 45/50w in gaming when is rated at 65w and well capable of drawing more? A bit odd.

65w == 1030 + the t rated i7.


Sent from my iPhone using Tapatalk

 

[Tapoer]

It might detect that the GPU is at 100% utilization and underclock the CPU since in the graphics the CPU is at a low clock. The power consumption of the 2400G is mostly the iGPU alone on witcher 3. On battlefield 1 in multiplayer the power consumption should be closer to the 65W TDP.

But yes all in all the 1030 might have a better performance per watt than the 2400G iGPU, but regardless Vega without high clocks have a decent perf/W.

 

[Qwertilot]

Clocking the cpu low then is sense. Wonder why the iGpu can't take more power though? Maybe the bandwidth throttles it enough that there's no point.

The perf/watt is definitely reasonable, just not a major win either


Sent from my iPhone using Tapatalk