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

Page 70 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
is it indicated in the product code/label?
so if it doesnt work it will usually be unstable as you go up, or chance of frying the memory, but not the memory controller right?

The cheaper modules on which the ICs used are not dictated by the specifications generally use whatever the manufacturer was able to source for the cheapest price at the time.
This means that it is not unusual that the ICs vary between the different batches. Some manufacturers have an identifier which specifies the memory IC model on the sticker, however the bulk modules generally don't.

Crucial mainstream models are always using Micron ICs, since Crucial is owned by Micron. There are some high-end Ballistix models which use Samsung ICs, since Micron themselves don't have anything available for high frequencies at competitive timings.
Kingston (ValueRam / HyperX) is generally Micron or Hynix based, only certain high-end HyperX models use Samsung ICs. Corsair meanwhile uses pretty much everything (Hynix, Micron / Spectek, Samsung), unless the ICs are dictated by the specifications.
G.Skill is nearly always either Hynix or Samsung, regardless of the model.

There won't be any physical damage to the hardware from memory overclocking, unless you increase the voltage to insane levels (on DDR4 1.5V or above).
 
Aug 11, 2008
10,451
642
126
Im pretty sure the market explosion of NUCs with Kaby Lage-G Chips will show otherwise.

4C/8T APU, with 1792 GCN5 cores, and 4 GB of HBM2 and 65-95W TDP will get TONS of traction. APU would consume everything up to GP107/GV107/Polaris 12, and to some degree Polaris 10 GPU market share. If properly designed, with heavy emphasize of front end(64 ROPs) it may even in performance cut into GTX 1060 3 GB levels.

I can assure you - insane amount of people would be interested in this chip.

P.S. If Full Raven Ridge is 2400G Ryzen 5, we still have room for bigger, more powerful APU in 2600G R5, and 2700G R7. Price target (199$ for 2600G and 299$ for 2700G) is also quite interesting from this perspective.
What explosion? The chip is not even released yet is it?
It could be a great product, but we still have to see price, power consumption, and performance relative to a discrete card to know if it will be more than a niche product. As for the monster APU, I will believe it when I see it. We have heard this speculation for years. Again lets see when or if it really comes to pass and what the pricing and performance is relative to discrete.
 
  • Like
Reactions: CHADBOGA

Eug

Lifer
Mar 11, 2000
23,583
996
126
So... ANYWAY, we should be cautious in assuming that Ryzen APUs will do Netflix 4K until it can be shown (proven).

:D
Reasonable point.

I had assumed 1030 would do Netflix 4K but it doesn’t.

There is also the matter of bugs or lack thereof, esp. early on.
 
  • Like
Reactions: nathanddrews

wahdangun

Golden Member
Feb 3, 2011
1,007
148
106
Why would AMD want to discourage people from buying their GPUs?

Again, it makes little financial sense for AMD to combine a high end CPU with a high end GPU in a monster APU. The market for that is miniscule.
No, it's not about that, but since miner make our life miserable with this gpu shortage, a monster APU can be some short of remedy.

And the monster Apu is already leaked, with 1750 shader and 2/4 GB hbm(pretty much Rx 570).
 
Aug 11, 2008
10,451
642
126
No, it's not about that, but since miner make our life miserable with this gpu shortage, a monster APU can be some short of remedy.

And the monster Apu is already leaked, with 1750 shader and 2/4 GB hbm(pretty much Rx 570).
Keep hearing about these "leaks". Do you have a source for it?
 

IRobot23

Senior member
Jul 3, 2017
601
183
76
Console killer APU?

Xbox one X has 6TFLOPs of compute power. Let's say that Consoles has 90-95% efficiency while PC has only around 65-55%. Xbox one X has lots of bandwidth.

Performance of Xbox one x is more like desktop gtx 1080 performance
 

wahdangun

Golden Member
Feb 3, 2011
1,007
148
106
Keep hearing about these "leaks". Do you have a source for it?
. Home News CPU, APU & Chipsets AMD's mysterious 'Fenghuang' APU teased with '15FF' graphics A new mysterious APU from AMD teased, known as 'Fenghuang' with '15FF' graphics By: Anthony Garreffa | CPU, APU & Chipsets News | Posted: Dec 14, 2017 6:16 am Comment | Email to a Friend | Font Size: AA AMD is launching multiple new CPUs and APUs next year but we're hearing about a new mysterious "Fenghuang" APU that has just appeared in the SiSoft Sandra database. AMD's upcoming Ryzen 2 processors will be using the company's Zen+ architecture and made on GlobalFroundries' 12LP node, but this new APU is something very interesting. The new "Fenghuang" APU will be powered by something called "15FF" graphics, which is said to rock 28 Radeon compute units (1792 shader processors) and 2GB of unknown-type VRAM. The APU is clocked at what looks to be a misidentified 555MHz, 16kB of L2 cache, and 32-bit path to memory. This new APU should rock GPU horsepower that is much faster than Vega 8 and Vega 10 IGPs, but we don't know how much power the APU will need yet.

Read more: https://www.tweaktown.com/news/60158/amds-mysterious-fenghuang-apu-teased-15ff-graphics/index.html
In sisoft it's 2 GB version but there will be 4 GB version also
 

whm1974

Diamond Member
Jul 24, 2016
9,460
1,570
96
No, it's not about that, but since miner make our life miserable with this gpu shortage, a monster APU can be some short of remedy.

And the monster Apu is already leaked, with 1750 shader and 2/4 GB hbm(pretty much Rx 570).
An APU with 1250 shaders I find hard to believe, especially if it has 8 corea and 16MB of L3 cache as that would be a fairly large die even at smaller processes. And wouldn't be cheap. A 6 core APU with 12 CUs and an 8MB L3 of cache along with 4GB of on package memory is far more realistic.
 

SPBHM

Diamond Member
Sep 12, 2012
5,056
409
126
Console killer APU?

Xbox one X has 6TFLOPs of compute power. Let's say that Consoles has 90-95% efficiency while PC has only around 65-55%. Xbox one X has lots of bandwidth.

Performance of Xbox one x is more like desktop gtx 1080 performance

I would compare the X more to the 1070 experience, and still, some games are locked to 30 due to CPU performance alone.

but even the original Xbox One can beat the 2400G (and GT 1030/RX 550), the specs are lower (apart from memory), but Witcher 3 is superior on the XB for example.
 

neblogai

Member
Oct 29, 2017
144
49
101
An APU with 1250 shaders I find hard to believe, especially if it has 8 corea and 16MB of L3 cache as that would be a fairly large die even at smaller processes. And wouldn't be cheap. A 6 core APU with 12 CUs and an 8MB L3 of cache along with 4GB of on package memory is far more realistic.

If using on-package memory- why not put on-package dGPU as well? There would be several advantages to this- like smaller (cheaper to manufacture) dies, and also the use of interchangeable dies for different market segments- while one large monolithic APU die can serve only one price level.
In this case- it would be reasonable to expect less iGPU resources on the APU die, not more. Just enough to serve with DDR4- when that same APU is packaged without a dGPU for budget segment, and also when low power operation at basic tasks is required.
And looking into the future- we can see AMD 'Horned Owl'/Raven Ridge at 15-65W on 14nm becoming 'Grey Hawk'/Picasso - same 4c/8t, on 7nm, and just 10-35W. So- no big APU die- but similar hardware shrunk and more efficient. I would expect this same APU to be reused in future AMD MCM products if they make them- same like Intel uses their quadcore APU for Kaby Lake G. Fenghuang River is also probably simply a name of a platform or package of Raven Ridge APU + Vega Mobile dGPU- not a huge APU by iself.
 

whm1974

Diamond Member
Jul 24, 2016
9,460
1,570
96
My thoughts on a 6 core APU with on package memory is something that AMD could sell for ~$200.
 

Timorous

Golden Member
Oct 27, 2008
1,532
2,535
136
So did a bit more tweaking last night as I was having stability issues with the RAM @ 3400 14-14-14-32 an have backed it off to 3400-16-15-15-36. Will tinker a bit more to see where the limit is. During that proess I backed of the GPU OC to stock to make sure that was not causing the issues so will be playing with that tonight.

I did boot up Stellaris at 4k and on the lowest settings it seems playable enough in my late game save.
 

whm1974

Diamond Member
Jul 24, 2016
9,460
1,570
96
So did a bit more tweaking last night as I was having stability issues with the RAM @ 3400 14-14-14-32 an have backed it off to 3400-16-15-15-36. Will tinker a bit more to see where the limit is. During that proess I backed of the GPU OC to stock to make sure that was not causing the issues so will be playing with that tonight.

I did boot up Stellaris at 4k and on the lowest settings it seems playable enough in my late game save.
That is nice to know that at least one game is actually playable at 4K with low settings on Raven Ridge. However I am a bit supprised though.
 

Timorous

Golden Member
Oct 27, 2008
1,532
2,535
136
That is nice to know that at least one game is actually playable at 4K with low settings on Raven Ridge. However I am a bit supprised though.

Well to me playable is the mouse moves smoothly and scrolling is smooth and it responds quickly to inputs, which it does. Framerates will tank when you have large fleets attacking each other but that happens no matter what hardware you use and it can be mitigated if you want. The thing is once fleets engage you are not really in control of them (emergency FTL aside) you can just watch it in slow mo and it makes no difference at all.

I expect the way I play CIV 6 will be playable at 4k too as I only use the 2d tactical map.
 
  • Like
Reactions: lightmanek

whm1974

Diamond Member
Jul 24, 2016
9,460
1,570
96
Well to me playable is the mouse moves smoothly and scrolling is smooth and it responds quickly to inputs, which it does. Framerates will tank when you have large fleets attacking each other but that happens no matter what hardware you use and it can be mitigated if you want. The thing is once fleets engage you are not really in control of them (emergency FTL aside) you can just watch it in slow mo and it makes no difference at all.

I expect the way I play CIV 6 will be playable at 4k too as I only use the 2d tactical map.
Not that I'm in the market for the 2400G, but if I was forced to build a new rig I might choose the APU as I refuse to pay the current high prices of dGPUs.
 
  • Like
Reactions: Peter Watts

Timorous

Golden Member
Oct 27, 2008
1,532
2,535
136
Not that I'm in the market for the 2400G, but if I was forced to build a new rig I might choose the APU as I refuse to pay the current high prices of dGPUs.

That is kind of where I am at, as well as rumours of the next gen NV GPUs coming out soon ish. Just feels like the 2200G is a good stop gap until I can get a better GPU. Although if it runs everything I throw at it well I may not even bother with a dGPU.
 
  • Like
Reactions: Peter Watts

del42sa

Member
May 28, 2013
26
11
81
IIRC AMD has said that Win 10 1709 is required for the APU to work properly.
So based on that I find Windows 7 support pretty unlikely, which would be a shame.

I have tried instal Raven Ridge APU on my windows 7 64bit, worked perfectly with Ryzen 5 1500X before. After bios update for RR support and swap Ryzen for RR system failed boot to windows..... I tried all SATA ports because I thought it wont boot because of M2 SSD disk which is connected directly to the CPU, so I made disc clone on sata SSD and connected it to the Sata ports from Promontory bridge. It doesnt help, system hangs a while after bios post when window animations starts. Then it boots again and said "required device is missing" "use installation disk and repair your computer". I tried everything but I was not able to load win7 system with RR. What is so different inside of this APU from Ryzen CPU ? Is there any way to make it work ? I dont really like win10. Could you test it Stilt please ? Maybe you will figure out, what is the problem. Note: with win 10 fresh installation APU is working flawlesly.
 

whm1974

Diamond Member
Jul 24, 2016
9,460
1,570
96
That is kind of where I am at, as well as rumours of the next gen NV GPUs coming out soon ish. Just feels like the 2200G is a good stop gap until I can get a better GPU. Although if it runs everything I throw at it well I may not even bother with a dGPU.
My problem is that since I already have a nice system, I would have to get at least 6 cores and an 1070 to see enough improvement to even bother with building a new rig.
 
  • Like
Reactions: Peter Watts

Peter Watts

Member
Jan 11, 2018
60
15
41
My problem is that since I already have a nice system, I would have to get at least 6 cores and an 1070 to see enough improvement to even bother with building a new rig.

You can wait for the new amd cpu's, that will supposedly come out in April.
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
I have tried instal Raven Ridge APU on my windows 7 64bit, worked perfectly with Ryzen 5 1500X before. After bios update for RR support and swap Ryzen for RR system failed boot to windows..... I tried all SATA ports because I thought it wont boot because of M2 SSD disk which is connected directly to the CPU, so I made disc clone on sata SSD and connected it to the Sata ports from Promontory bridge. It doesnt help, system hangs a while after bios post when window animations starts. Then it boots again and said "required device is missing" "use installation disk and repair your computer". I tried everything but I was not able to load win7 system with RR. What is so different inside of this APU from Ryzen CPU ? Is there any way to make it work ? I dont really like win10. Could you test it Stilt please ? Maybe you will figure out, what is the problem. Note: with win 10 fresh installation APU is working flawlesly.

I've seen this as well.
I think it is just the SATA drivers, but I need to check.
 

Timorous

Golden Member
Oct 27, 2008
1,532
2,535
136
My problem is that since I already have a nice system, I would have to get at least 6 cores and an 1070 to see enough improvement to even bother with building a new rig.

In my case I was upgrading from an old laptop with a 540m so even intel graphics was an upgrade. CPU less so as it did have a 4c 8t sandy but low clocks.
 

cbn

Lifer
Mar 27, 2009
12,968
221
106
No, it's not about that, but since miner make our life miserable with this gpu shortage, a monster APU can be some short of remedy.

And the monster Apu is already leaked, with 1750 shader and 2/4 GB hbm(pretty much Rx 570).

Keep hearing about these "leaks". Do you have a source for it?

Console killer APU?

Xbox one X has 6TFLOPs of compute power. Let's say that Consoles has 90-95% efficiency while PC has only around 65-55%. Xbox one X has lots of bandwidth.

Performance of Xbox one x is more like desktop gtx 1080 performance

wahdungun said:
Home News CPU, APU & Chipsets AMD's mysterious 'Fenghuang' APU teased with '15FF' graphics A new mysterious APU from AMD teased, known as 'Fenghuang' with '15FF' graphics By: Anthony Garreffa | CPU, APU & Chipsets News | Posted: Dec 14, 2017 6:16 am Comment | Email to a Friend | Font Size: AA AMD is launching multiple new CPUs and APUs next year but we're hearing about a new mysterious "Fenghuang" APU that has just appeared in the SiSoft Sandra database. AMD's upcoming Ryzen 2 processors will be using the company's Zen+ architecture and made on GlobalFroundries' 12LP node, but this new APU is something very interesting. The new "Fenghuang" APU will be powered by something called "15FF" graphics, which is said to rock 28 Radeon compute units (1792 shader processors) and 2GB of unknown-type VRAM. The APU is clocked at what looks to be a misidentified 555MHz, 16kB of L2 cache, and 32-bit path to memory. This new APU should rock GPU horsepower that is much faster than Vega 8 and Vega 10 IGPs, but we don't know how much power the APU will need yet.

Read more: https://www.tweaktown.com/news/60158/amds-mysterious-fenghuang-apu-teased-15ff-graphics/index.html

In sisoft it's 2 GB version but there will be 4 GB version also

16kB of L2 cache? How many cpu cores would that be on 12nm?
 
May 11, 2008
19,300
1,129
126
The cheaper modules on which the ICs used are not dictated by the specifications generally use whatever the manufacturer was able to source for the cheapest price at the time.
This means that it is not unusual that the ICs vary between the different batches. Some manufacturers have an identifier which specifies the memory IC model on the sticker, however the bulk modules generally don't.

Crucial mainstream models are always using Micron ICs, since Crucial is owned by Micron. There are some high-end Ballistix models which use Samsung ICs, since Micron themselves don't have anything available for high frequencies at competitive timings.
Kingston (ValueRam / HyperX) is generally Micron or Hynix based, only certain high-end HyperX models use Samsung ICs. Corsair meanwhile uses pretty much everything (Hynix, Micron / Spectek, Samsung), unless the ICs are dictated by the specifications.
G.Skill is nearly always either Hynix or Samsung, regardless of the model.

There won't be any physical damage to the hardware from memory overclocking, unless you increase the voltage to insane levels (on DDR4 1.5V or above).

Does this explain why the mainstream crucial modules are widely available ?
I have seen 2666MHZ@1.2V mainstream models available but when i check the US,and UK sites of crucial ballistix and ballistix elite , most is out of stock.
I always noticed that the ballistix modules run standard on 2666MHz @1.2V as well as advertised on the crucial ballistix site. I assumed that it was the 2666MHz DDR4 micron chips.

Other question : Why do so many modules from other manufacturers start default at 2133MHz @1.2V but run at 3200MHz@1.35V ?
Is that not a huge leap in frequency ?
 

The Stilt

Golden Member
Dec 5, 2015
1,709
3,057
106
Does this explain why the mainstream crucial modules are widely available ?
I have seen 2666MHZ@1.2V mainstream models available but when i check the US,and UK sites of crucial ballistix and ballistix elite , most is out of stock.
I always noticed that the ballistix modules run standard on 2666MHz @1.2V as well as advertised on the crucial ballistix site. I assumed that it was the 2666MHz DDR4 micron chips.

Other question : Why do so many modules from other manufacturers start default at 2133MHz @1.2V but run at 3200MHz@1.35V ?
Is that not a huge leap in frequency ?

The Ballistix products featuring non-Micron ICs are most likely a niche / filler product to complete the lineup to Crucial, and therefore the production is limited as well.

DRAM IC manufacturers follow JEDEC specifications for their products and the nominal supply voltage for normal DDR4 is 1.20V.
3200MHz at 1.20V rated ICs have recently entered into production or sampling phase, but the low supply voltage combined at high speed comes at a cost of extremely high latencies (e.g. 22-22-22 tCL-tRCD-tRP for Hynix 3200MHz ICs).

Most of the recent DDR4 ICs can be overclocked to far above (>= 3000MHz) their rated speeds, but generally that requires increasing the voltage significantly higher than the JEDEC specification dictates (1.35V or higher).