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

Page 29 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

PeterScott

Platinum Member
Jul 7, 2017
2,605
1,540
106
My prediction: Ryzen 5 2400G will perform 20% slower than RX 460 with 3600 memory and 1675MHz overclock (as achieved in the charts).

This puts it roughly 30% behind desktop GTX 1050 depending on game.
I think the more reasonable comparison is RX 550, and even overclocked it will struggle to meet parity with that.

RX 550 has 112 GB/s of bandwidth. If AMD didn't feel they benefit from the extra bandwidth, why would they put a more expensive 128 bit DD4 bus on the budget part?

I am not saying they need 112 GB/s for RX550, but there is a strong case that half (56 GB/s) wasn't considered to be enough.

RX 550/GT1030 can be viewed a reasonable ceiling on the RR GPU performance.
 
  • Like
Reactions: tential

Yotsugi

Golden Member
Oct 16, 2017
1,029
487
106
If AMD didn't feel they benefit from the extra bandwidth, why would they put a more expensive 128 bit DD4 bus on the budget part?
Inevitable DDR4 version of 550? Call it 540 or something.
If HBM is to be added to APU i would do it externally, petty much like DDR3 sideport on DDR2 era, let the user decide that.
Um what?
End user is gonna do wafer-level integration?
 

Shivansps

Diamond Member
Sep 11, 2013
3,641
1,329
136
To be clear on this (again) the AMD own slides are the ones that makes me suspect something is wrong.


And this one


Those numbers, for 1080p @ low are kinda low even for a G4560+GT1030, and the 2400G has a huge advantage on CPU side here.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,574
126
To be clear on this (again) the AMD own slides are the ones that makes me suspect something is wrong.


And this one


Those numbers, for 1080p @ low are kinda low even for a G4560+GT1030, and the 2400G has a huge advantage on CPU side here.
Well, the second one has to be just using HD630. I don't think anyone expected the Intel IGP to keep up with the AMD APU graphics. They never have, and it's never made much difference.

The first one shows a tie with a GT1030, and that may be a best case slide. Or it may not.
It is also emphasizing the power use. So it appears to basically be claiming the same performance as a GT1030 while using less power.
 

CatMerc

Golden Member
Jul 16, 2016
1,114
1,149
136
I think the more reasonable comparison is RX 550, and even overclocked it will struggle to meet parity with that.

RX 550 has 112 GB/s of bandwidth. If AMD didn't feel they benefit from the extra bandwidth, why would they put a more expensive 128 bit DD4 bus on the budget part?

I am not saying they need 112 GB/s for RX550, but there is a strong case that half (56 GB/s) wasn't considered to be enough.

RX 550/GT1030 can be viewed a reasonable ceiling on the RR GPU performance.
64bit bus would limit the card to 2GB VRAM.
 

Yotsugi

Golden Member
Oct 16, 2017
1,029
487
106
They should add the controller but not the memory, let it be a motherboard addon, just like sideport was.
You do understand how HBM even works?
It uses really wide interface.
So wide no PCB could be this dense.
 

Shivansps

Diamond Member
Sep 11, 2013
3,641
1,329
136
Well, the second one has to be just using HD630. I don't think anyone expected the Intel IGP to keep up with the AMD APU graphics. They never have, and it's never made much difference.
What i meant, is that some of those numbers are a bit low compared to a 1030, is not easy to compare because it all depends on the testing method, but an GT1030 on BF1@low@1080p gets about 75fps on avg and the Witcher 3 runs on Medium at about 30-35 fps.

maybe is the drivers or the memory used, but considering this is an AMD slide, i would consider this the best case escenario.
They avoided a direct gaming comparison with the 1030. That much is clear.
 

CatMerc

Golden Member
Jul 16, 2016
1,114
1,149
136
Then GDDR5/GDDR6 interfance. Increasing APU price by adding HBM is not an option.


That is a non-issue on a RX550.
You see it as a non issue, AMD doesn't. They see it as a value add over the competing 2GB cards. It's classic AMD to load up their cards with more VRAM than their competition whenever possible.
 
  • Like
Reactions: DarthKyrie

LTC8K6

Lifer
Mar 10, 2004
28,520
1,574
126
You see it as a non issue, AMD doesn't. They see it as a value add over the competing 2GB cards. It's classic AMD to load up their cards with more VRAM than their competition whenever possible.
RX550 comes in a 2gb version anyway...
 

Shivansps

Diamond Member
Sep 11, 2013
3,641
1,329
136
You see it as a non issue, AMD doesn't. They see it as a value add over the competing 2GB cards. It's classic AMD to load up their cards with more VRAM than their competition whenever possible.
4GB RX550 is petty much a scam.
 

Abwx

Diamond Member
Apr 2, 2011
9,814
2,247
136

Source are voltage/frequency curves available for 14nm LPP used for current products, you can use the one posted by The Stilt elsewhere.

Power delta = frequency delta x (voltage delta)^2

So far for 14nm LPP the voltage delta applied to a given frequency delta is such that :

Power delta = (frequency delta)^2.56

If the process is improved such that it allow 11% higher frequency at same power (they said 10%+...) then it means that at same frequency the power delta will be 1.1^2.56 = 1.306x lower..

Actually the most interesting is wether they managed to keep this rate significantly above 3.3GHz, wich is the frequency above wich the current process diverge from this rate and enter a cubic and even quadratic rate, if they improved max frequency by say 10% then all CPUs clocked above 3.4 will benefit from an extra improvement on top of said 30% within this range.

but an GT1030 on BF1@low@1080p gets about 75fps on avg and the Witcher 3 runs on Medium at about 30-35 fps.
.
 
Last edited:

piesquared

Golden Member
Oct 16, 2006
1,651
473
136
Well, the second one has to be just using HD630. I don't think anyone expected the Intel IGP to keep up with the AMD APU graphics. They never have, and it's never made much difference.

The first one shows a tie with a GT1030, and that may be a best case slide. Or it may not.
It is also emphasizing the power use. So it appears to basically be claiming the same performance as a GT1030 while using less power.
You are not thinking that line of thought through very well. Prior to Ryzen APUs, these were basically the options:

1). Buy an AMD APU, which currently is Bristol Ridge. Although with an excellent integrated GPU and efficiency, it lacked the CPU grunt to compete under the reality of competing market conditions in general computing.

2). Buy an intel APU with the compute grunt on the CPU side and suffer the consequences of using an intel GPU.

3). Buy an intel APU and attach a discrete GPU.

4). Buy an AMD APU and attach a discrete GPU.


And so from those options it's clear to see why AMD's strong GPU in construction core APUs wouldn't make much of a difference. Consumers and OEMs were buying the intel APU and having to attach a discrete GPU if they wanted to be able to do anything with satisfactory results besides watching a video. If playing a game wasn't a requirement, consumers and OEMs were also buying intel APUs, again because of the compute performance.

Obviously it is a completely different paradigm with Ryzen APUs, since as everyone knows, they have virtually matched the compute performance of the competition. Consumers now have the viable option of purchasing a Ryzen APU, or purchase the intel APU and attach an entry level GPU for the same performance at higher cost, higher TDP and poorer software compared to Radeon Settings. In the case of an intel + Nvidia combination, that also means multiple drivers and software packages to manage.
 

LTC8K6

Lifer
Mar 10, 2004
28,520
1,574
126
You are not thinking that line of thought through very well. Prior to Ryzen APUs, these were basically the options:

1). Buy an AMD APU, which currently is Bristol Ridge. Although with an excellent integrated GPU and efficiency, it lacked the CPU grunt to compete under the reality of competing market conditions in general computing.

2). Buy an intel APU with the compute grunt on the CPU side and suffer the consequences of using an intel GPU.

3). Buy an intel APU and attach a discrete GPU.

4). Buy an AMD APU and attach a discrete GPU.


And so from those options it's clear to see why AMD's strong GPU in construction core APUs wouldn't make much of a difference. Consumers and OEMs were buying the intel APU and having to attach a discrete GPU if they wanted to be able to do anything with satisfactory results besides watching a video. If playing a game wasn't a requirement, consumers and OEMs were also buying intel APUs, again because of the compute performance.

Obviously it is a completely different paradigm with Ryzen APUs, since as everyone knows, they have virtually matched the compute performance of the competition. Consumers now have the viable option of purchasing a Ryzen APU, or purchase the intel APU and attach an entry level GPU for the same performance at higher cost, higher TDP and poorer software compared to Radeon Settings. In the case of an intel + Nvidia combination, that also means multiple drivers and software packages to manage.
I don't think the paradigm has changed much. While the AMD APUs are certainly better than they were, I don't think the overall situation will change much.
I'll be surprised if in 6 months, Intel's IGP has lost much video market share.
 

CatMerc

Golden Member
Jul 16, 2016
1,114
1,149
136
Source are voltage/frequency curves available for 14nm LPP used for current products, you can use the one posted by The Stilt elsewhere.

Power delta = frequency delta x (voltage delta)^2

So far for 14nm LPP the voltage delta applied to a given frequency delta is such that :

Power delta = (frequency delta)^2.56

If the process is improved such that it allow 11% higher frequency at same power (they said 10%+...) then it means that at same frequency the power delta will be 1.1^2.56 = 1.306x lower..

Actually the most interesting is wether they managed to keep this rate significantly above 3.3GHz, wich is the frequency above wich the current process diverge from this rate and enter a cubic and even quadratic rate, if they improved max frequency by say 10% then all CPUs clocked above 3.4 will benefit from an extra improvement on top of said 30% within this range.
Ah, I was confused because I incorrectly remembered the 15% density improvement as 15% improvement to efficiency. My bad.
 

piesquared

Golden Member
Oct 16, 2006
1,651
473
136
I don't think the paradigm has changed much. While the AMD APUs are certainly better than they were, I don't think the overall situation will change much.
I'll be surprised if in 6 months, Intel's IGP has lost much video market share.
I think it has. If intel's IGP *doesn't lose much video market share, then that would point to something else going on in the market (shades of historical anti-competitive practices for example), because as far as capability, efficiency and performance, Ryzen APUs have leveled the playing field while continuing to obliterate the competitions IGP. Plus AMD isn't sullied with the stigma of intel's Meltdown problem.

*changed hasn't to doesn't
 
Last edited:
  • Like
Reactions: DarthKyrie

LTC8K6

Lifer
Mar 10, 2004
28,520
1,574
126
I think it has. If intel's IGP hasn't lost much video market share, then that would point to something else going on in the market (shades of historical anti-competitive practices for example), because as far as capability, efficiency and performance, Ryzen APUs have leveled the playing field while continuing to obliterate the competitions IGP. Plus AMD isn't sullied with the stigma of intel's Meltdown problem.
Ryzen APUs haven't done a single thing yet. They have not been launched.

Intel still has most of the graphics market, despite years of AMD APUs that crush Intel's IGP offerings.

https://cdn.wccftech.com/wp-content/uploads/2017/11/GPU-Market-Share-Q3-2017-NVIDIA-Intel-AMD.png
 

piesquared

Golden Member
Oct 16, 2006
1,651
473
136
Ryzen APUs haven't done a single thing yet. They have not been launched.

Intel still has most of the graphics market, despite years of AMD APUs that crush Intel's IGP offerings.

https://cdn.wccftech.com/wp-content/uploads/2017/11/GPU-Market-Share-Q3-2017-NVIDIA-Intel-AMD.png
You are not thinking that line of thought through very well. Prior to Ryzen APUs, these were basically the options:

1). Buy an AMD APU, which currently is Bristol Ridge. Although with an excellent integrated GPU and efficiency, it lacked the CPU grunt to compete under the reality of competing market conditions in general computing.

2). Buy an intel APU with the compute grunt on the CPU side and suffer the consequences of using an intel GPU.

3). Buy an intel APU and attach a discrete GPU.

4). Buy an AMD APU and attach a discrete GPU.


And so from those options it's clear to see why AMD's strong GPU in construction core APUs wouldn't make much of a difference. Consumers and OEMs were buying the intel APU and having to attach a discrete GPU if they wanted to be able to do anything with satisfactory results besides watching a video. If playing a game wasn't a requirement, consumers and OEMs were also buying intel APUs, again because of the compute performance.

Obviously it is a completely different paradigm with Ryzen APUs, since as everyone knows, they have virtually matched the compute performance of the competition. Consumers now have the viable option of purchasing a Ryzen APU, or purchase the intel APU and attach an entry level GPU for the same performance at higher cost, higher TDP and poorer software compared to Radeon Settings. In the case of an intel + Nvidia combination, that also means multiple drivers and software packages to manage.
 
Aug 11, 2008
10,451
642
126
I don't think the paradigm has changed much. While the AMD APUs are certainly better than they were, I don't think the overall situation will change much.
I'll be surprised if in 6 months, Intel's IGP has lost much video market share.
Yea, I agree. The problem with AMD's previous Apus was not really the cpu, it was the bandwidth limited igpu. Granted it was better than intel's, but compared to a discrete card, it was like bringing a slightly bigger knife to a gunfight. That hasnt really changed much. We wont know until we see real benchmarks, but the apu almost certainly will still struggle to match even a low end discrete card. It is the same problem that AMD has always had. If you dont game, either AMD or Intel igpus are good enough. If you do game, a discrete card will give a much better experience.
 
Aug 11, 2008
10,451
642
126
I think it has. If intel's IGP *doesn't lose much video market share, then that would point to something else going on in the market (shades of historical anti-competitive practices for example), because as far as capability, efficiency and performance, Ryzen APUs have leveled the playing field while continuing to obliterate the competitions IGP. Plus AMD isn't sullied with the stigma of intel's Meltdown problem.

*changed hasn't to doesn't
Wow, that kind of "logic" must be very comforting. If your team wins, all is good. If they dont, it is because the other team cheated. Or maybe there is a third option: the vast majority of buyers dont really know or care much about how strong an igpu is, and the enthusiasts who do care will buy a discrete card.
 
  • Like
Reactions: tential

piesquared

Golden Member
Oct 16, 2006
1,651
473
136
Wow, that kind of "logic" must be very comforting. If your team wins, all is good. If they dont, it is because the other team cheated. Or maybe there is a third option: the vast majority of buyers dont really know or care much about how strong an igpu is, and the enthusiasts who do care will buy a discrete card.
Especially if the team in question has a long history of doing just that. All else being equal, precedence is a legitimate factor.

It must not be comforting to have to move the goal posts all over the field to make your point possible.

The problem with AMD's previous Apus was not really the cpu, it was the bandwidth limited igpu.
Or maybe there is a third option: the vast majority of buyers dont really know or care much about how strong an igpu is, and the enthusiasts who do care will buy a discrete card.
And you inadvertently made my point in your last statement. If the vast majority of buyers don't care much about how strong an igpu is, then it stands to reason that they care more about compute performance and is what was holding back sales of Bristol, Carrizo, Kaveri and Trinity. Ryzen APUs eliminates that deficit, and gives buyers the ability to have discrete class graphics paired with strong CPU cores without needing the discrete graphics and the complexity of multiple drivers/software packages from different vendors. Buyers were attaching discrete GPUs to intel APUs because they had no choice if they wanted anything but a failed gaming experience.
 
Last edited:

ASK THE COMMUNITY