AMD Ryzen 5 2400G and Ryzen 3 2200G APUs performance unveiled

[frozentundra123456]

Results are starting trickle out. Add another 30% from overclocking and this thing could approach 1050/ RX 560 territory! Beast.

http://moepc.net/?post=4249

Ryzen Mobile @ 15W is amazing also.

Hold on there. Better check the math. GTX 1050 is 70% faster than the 2400G. *If* you can overclock the 2400 30%, and *if* the performance scales linearly with clockspeed, and *if* you dont run into thermal, TDP or bandwidth limitations with overclocking, that would give the 2400 a score of 6500. The 1050 is still more than 30% faster, and the 1050 can be overclocked as well.

 

[neblogai]

In test setup he had DDR4-2933 for Raven Ridge APUs.

If you are referring to http://moepc.net/?post=4249 - that is not a test setup, it only has a table with 2400G and 2200G official specifications. After that- the score tables are made up out of anonymous 3DMark results discovered by Tum_Apisak. No tests were done by the author of the article- it's just an article made up out of tidbits of information from leaks.

 

[SPBHM]

we have official scores for Firestrike, so it's probably better to speculate on that? the 3dm11 can be real, probably are, but we don't have an idea if it's stock or OC and memory speed

https://www.pcper.com/files/news/2018-01-07/08.jpg

stock is around 2900-3300 (they didn't test with 2933 ram)
we can compare that with
https://us.hardware.info/reviews/76...ncake-benchmarksn3dmark-fire-strike-ntime-spy

GTX 1050 = 6700
GT 1030 = 3500
2400G = 2900-3300
2400G OC = 4000

I found from a different source that the 750 Ti stock is around 4000.

but again, 3dmark tends to be more forgiving with the APUs memory bandwidth than actual games on average.

 

[nickmania]

we need DDR5 for this chips and also an AV1 decode block to be the perfect purchase to last years.

 

[Sane Indian]

AMD should have added 1-4 GB DDR5 as unified L4 cache for APUs. AMD has already implemented that in PS4 and X Box one. It would have addressed memory bottle necks of iGPU and it would have bumped up even CPU's performance.

 

[itsmydamnation]

AMD should have added 1-4 GB DDR5 as unified L4 cache for APUs. AMD has already implemented that in PS4 and X Box one. It would have addressed memory bottle necks of iGPU and it would have bumped up even CPU's performance.

that is a terrible idea, you would smash CPU performance, understand what making something a cache actually means within a memory hierarchy.

Far better would be a "side port" of GDDR5 and then rely on the HBCC to make use of it.

 

[IRobot23]

An RX 550 has the same mem bandwidth and ROPs as an RX 560, and the 3DMark 11P score divided by FLOPs is 15-20% higher than that of the 560. Being overprovisioned with bandwidth and ROPs, the RX 550 is very much an outlier for GCN GPUs, and thus a naive comparison of any other GCN GPU to an RX 550 will give inaccurate conclusions.

VEGA is different by ROPs.
We can expect 16 ROPs, 1 shader engine (graphic pipeline) vs 2 shader engines RX 550, but in this case with fast DRAM maybe you can approach ~55GB/s (copy@AIDA64) which is double of previous APU with fast DDR3 and ~2/3 of RX 560.



Distribution of memory bandwidth could be much better in this case since new technology (IF). Yet CPU will take some bandwidth.

 

[epsilon84]

we have official scores for Firestrike, so it's probably better to speculate on that? the 3dm11 can be real, probably are, but we don't have an idea if it's stock or OC and memory speed

https://www.pcper.com/files/news/2018-01-07/08.jpg

stock is around 2900-3300 (they didn't test with 2933 ram)
we can compare that with
https://us.hardware.info/reviews/76...ncake-benchmarksn3dmark-fire-strike-ntime-spy

GTX 1050 = 6700
GT 1030 = 3500
2400G = 2900-3300
2400G OC = 4000

I found from a different source that the 750 Ti stock is around 4000.

but again, 3dmark tends to be more forgiving with the APUs memory bandwidth than actual games on average.

Just ran 3DMark again out of curiosity on my 1050 rig (its been a while) and got ~6800 stock / ~7600 OC for the graphics scores in Firestrike.

If the 2400G can get ~50% the gaming performance of a 1050 I would be impressed, as that would make it a viable 1080P gaming solution even on AAA titles, albeit at low settings, which is unheard of for an iGPU. I suspect it would be closer to 30 - 40% that of a 1050 though in actual gaming.

 

[raghu78]

Just ran 3DMark again out of curiosity on my 1050 rig (its been a while) and got ~6800 stock / ~7600 OC for the graphics scores in Firestrike.

If the 2400G can get ~50% the gaming performance of a 1050 I would be impressed, as that would make it a viable 1080P gaming solution even on AAA titles, albeit at low settings, which is unheard of for an iGPU. I suspect it would be closer to 30 - 40% that of a 1050 though in actual gaming.

https://www.pcper.com/image/view/88653?return=node/69029

the 2400G firestrike numbers are from official AMD presentation at CES Tech day 2018. So definitely 3300 - 3500 is going to be the range for R5 2400G with DDR4 3200 and a slight bump in clocks. GT1030 class iGPU using just system DDR4 is amazing.

I have been saying this for years and I stand by it. The APU in the long run will obsolete the 100-120 sq mm dGPU. From 7nm process onwards I think AMD will add a HBM cache and deliver perf on par with $150 dGPU, completely making the <= $150 dGPU redundant. All that remains is for HBM2 memory to become more cost effective and packaging technologies to become more cost effective. I think in a couple of years we will reach that stage. I think the 7nm APUs which will launch in early 2020 are going to be the complete realization of AMD's Fusion and Vision 25x20 goal.

 

[CatMerc]

we have official scores for Firestrike, so it's probably better to speculate on that? the 3dm11 can be real, probably are, but we don't have an idea if it's stock or OC and memory speed

https://www.pcper.com/files/news/2018-01-07/08.jpg

stock is around 2900-3300 (they didn't test with 2933 ram)
we can compare that with
https://us.hardware.info/reviews/76...ncake-benchmarksn3dmark-fire-strike-ntime-spy

GTX 1050 = 6700
GT 1030 = 3500
2400G = 2900-3300
2400G OC = 4000

I found from a different source that the 750 Ti stock is around 4000.

but again, 3dmark tends to be more forgiving with the APUs memory bandwidth than actual games on average.

We don't know if those results are overall firestrike score or graphics score. Therefore direct comparisons have a wildly large margin of error.

 

[Shivansps]

We don't know if those results are overall firestrike score or graphics score. Therefore direct comparisons have a wildly large margin of error.

Petty sure its general score, 3300 general score its petty much spot on with R5 1400 + GT1030, and a bit worse if its graphics only. We need to consider the R5 1400 is slower than 2400G CPU only as well.
This is what AMD has been telling so far, a Timespy result equal to a I5-8400 + GT1030, the thing is, there is no I5-8400+GT1030 timespy result in the 3dmark database, somehow no one has been dumb enoght to pair a 8400 with a GT 1030 and upload a result. There is a single Firestrike result... so this is kinda an imaginary enemy AMD invented here.

Anyway, everything seems to point Stock+DDR4-3200 equal to GT1030 in 3dmark for the 2400G, what would translate to slower than GT1030 in games. I really hope is not that way because that would be the worse APU to dGPU perf ratio in the history of APUs.

 

[alexruiz]

Less than 2 weeks to go!
I would have expected more and more benchmarks to start leaking

 

[raghu78]

Petty sure its general score, 3300 general score its petty much spot on with R5 1400 + GT1030, and a bit worse if its graphics only. We need to consider the R5 1400 is slower than 2400G CPU only as well.
This is what AMD has been telling so far, a Timespy result equal to a I5-8400 + GT1030, the thing is, there is no I5-8400+GT1030 timespy result in the 3dmark database, somehow no one has been dumb enoght to pair a 8400 with a GT 1030 and upload a result. There is a single Firestrike result... so this is kinda an imaginary enemy AMD invented here.

Anyway, everything seems to point Stock+DDR4-3200 equal to GT1030 in 3dmark for the 2400G, what would translate to slower than GT1030 in games. I really hope is not that way because that would be the worse APU to dGPU perf ratio in the history of APUs.

How did you come to the conclusion that 2400G will be slower in games compared to GT 1030 ? Anyway we will know in exactly 10-11 days time what the 2400G can do. Rest assured the 2400G and 2200G are going to be extremely popular in budget PCs.

 

[Yotsugi]

How did you come to the conclusion that 2400G will be slower in games compared to GT 1030 ? Anyway we will know in exactly 10-11 days time what the 2400G can do. Rest assured the 2400G and 2200G are going to be extremely popular in budget PCs.

2200G would've been an absolute marvel (4/4 + iGPU with actual drivers for $99) if not the memory prices.

 

[Abwx]

that would be the worse APU to dGPU perf ratio in the history of APUs.

What about the APU to CPU ratio..?

If CPU ratio is that much upped it s no wonder that the APU to dGPU ratio is less favourable even if the whole APU was 10x faster..

Here the CPU has 2.5x the throughput of Bristol Ridge while the GPU has "only" 55% or so more Gflops, that s what you call worse, i guess that the previous complaint from a given public was that the CPU wasnt up to the GPU, now that you have i5/i7 perf level with a decent GPU at 50/60% of the price we are told that it s the worse APU ever, lol...

FTR 70% of DTs have only an iGPU, so the debate is not wether a CPU+dGPU would be better but what is the best APU for 70% of the DT market..

 

[raghu78]

2200G would've been an absolute marvel (4/4 + iGPU with actual drivers for $99) if not the memory prices.

2200G is a fantastic APU inspite of memory prices. The price difference between dual channel DDR4 2400 kit and dual channel DDR4 3200 kit is $17 on newegg.

https://www.newegg.com/Product/Product.aspx?Item=N82E16820232608

https://www.newegg.com/Product/Product.aspx?Item=N82E16820231900

Thats roughly the same price diff between 2200G and 8100. So for the same price of APU+ memory you get a 2200G which is probably 10% slower in ST than 8100 and comparable on MT but twice as fast in games and has an unlocked multiplier to further close the ST gap. Phenomenal value.

 

[Yotsugi]

2200G is a fantastic APU inspite of memory prices.

That would've been even better if memory prices were not that bad.

 

[LTC8K6]

You must have two sticks of DDR4 with the APUs or the graphics bandwidth will be bad, but with a GT1030 or RX550 and an R5-1400, you can use just one stick of DDR4 without suffering much.

I think overall, the value of the 2400G will be questionable.
The 2200G looks better.

Soon we should have an R5-2400 chip that is a bit faster and a bit cheaper than the R5-1400.

Maybe by then video card prices will have returned to sanity.

 

[Abwx]

I think overall, the value of the 2400G will be questionable.

Why is it questionable, because it s AMD ?

I mean, was there a competing product that was , or is, competitive, say an i7 for instance, was there one at 169$ that performed better than the 2400G ?...

 

[IntelUser2000]

2200G is really nice for the price. Technically, even the cheapest memory would work fine, to maximize value out of the money spent. Personally, I don't like messing around with RAM settings, so I would get speeds I need rather than OC them. Even at DDR4-2133, the iGPU would perform admirably. Though it costs minimally to get DDR4-2400, or even DDR4-2666 modules.

You must have two sticks of DDR4 with the APUs or the graphics bandwidth will be bad, but with a GT1030 or RX550 and an R5-1400, you can use just one stick of DDR4 without suffering much.

That's true, but you'd still lose quite a bit of system performance. 10-20% is expected. That's the cheapest 10-20% boost you can give to your system.

 

[piesquared]

What about the APU to CPU ratio..?

If CPU ratio is that much upped it s no wonder that the APU to dGPU ratio is less favourable even if the whole APU was 10x faster..

Here the CPU has 2.5x the throughput of Bristol Ridge while the GPU has "only" 55% or so more Gflops, that s what you call worse, i guess that the previous complaint from a given public was that the CPU wasnt up to the GPU, now that you have i5/i7 perf level with a decent GPU at 50/60% of the price we are told that it s the worse APU ever, lol...

FTR 70% of DTs have only an iGPU, so the debate is not wether a CPU+dGPU would be better but what is the best APU for 70% of the DT market..

Well, there are basically less than a handful of people are claiming that it is the worst APU ever so don't take those claims to seriously, it just marketing fluff. If that were the case it would make competing APUs in that class completely useless and irrelevant. Ryzen G series would have similar single threaded performance, multi threaded performance and far and away better graphics performance. And with one stick of RAM it is all that plus even lower cost.

 

[piesquared]

You must have two sticks of DDR4 with the APUs or the graphics bandwidth will be bad, but with a GT1030 or RX550 and an R5-1400, you can use just one stick of DDR4 without suffering much.

I think overall, the value of the 2400G will be questionable.
The 2200G looks better.

Soon we should have an R5-2400 chip that is a bit faster and a bit cheaper than the R5-1400.

Maybe by then video card prices will have returned to sanity.

With a Ryzen G series, performance with one stick of RAM would still be enough to handily outperform competing APUs. For more performance, add another stick and it would still be cheaper with similar performance to a 1030. No matter which way you slice it, Ryzen G series is a win.

 

[neblogai]

Why is it questionable, because it s AMD ?

I mean, was there a competing product that was , or is, competitive, say an i7 for instance, was there one at 169$ that performed better than the 2400G ?...

It is questionable when you don't need the iGPU. Simply as a CPU- it is too close to the price of R5 1600 and i5 8400 (with H310).
But if you need the iGPU, and 4c/8t is what you need from CPU- then 2400G is very well priced.

 

[Yotsugi]

But if you need the iGPU, and 4c/8t is what you need from CPU- then 2400G is very well priced.

The very point of any "G" SKU is to specifically adress the "I need iGPU" niche.
Even then, 2200G still seems like a better value.

 

[LTC8K6]

Why is it questionable, because it s AMD ?

I mean, was there a competing product that was , or is, competitive, say an i7 for instance, was there one at 169$ that performed better than the 2400G ?...

Yes, that's why I suggested an R5-1400 and a DGPU and not an Intel chip...because I don't like AMD...