Zen 2 APUs/"Renoir" discussion thread

Page 5 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Asterox

Senior member
May 15, 2012
331
322
136
Now i will sit in ambush, and wait for Ryzen 3 Pro 4350GE.

F51f is new Gigabyte B450 Bios with suport for Renoir APU.


I checked at random, all Gigabyte older AM4 motherboards(even A320 models) has new Bios with suport for Renoir APU.

 
Last edited:

grant2

Golden Member
May 23, 2001
1,153
16
81
Now i will sit in ambush, and wait for Ryzen 3 Pro 4350GE.
What's the advantage of the GE variant over the G variant?

Obviously power consumption. But isn't it possible to set a power profile in your motherboard that also limits to 35w TDP without needing a unique model for that?
 

Asterox

Senior member
May 15, 2012
331
322
136
What's the advantage of the GE variant over the G variant?

Obviously power consumption. But isn't it possible to set a power profile in your motherboard that also limits to 35w TDP without needing a unique model for that?
Maybe or maybe not, or a lower priced 35W TDP version has fixed TDP+no overclocking.

If we compare old and new 65W APU, Renoir 6 CU is not that much slower vs "standard Vega 11 CU".

4/8 R5 3400G, 65W

- 11 CU GPU score 1322

- CPU score 4200


4/8 R3 Pro 4350G, 65W

2020-07-12_131525.jpg



 

grant2

Golden Member
May 23, 2001
1,153
16
81
If a 35w version is cheaper than 65w, then i understand why it's worth buying.

But if a person is searching for a specific power envelope, and they're confident with their mobo setup skills, then it looks like getting the 65w version leaves their options open in case they ever change their mind.
 

Shivansps

Platinum Member
Sep 11, 2013
2,914
631
136
Wait wait wait... AMD is planing to price-replace the 3400G with a 4300G? This is in no way shape or form acceptable.
 

Shivansps

Platinum Member
Sep 11, 2013
2,914
631
136
If we compare old and new 65W APU, Renoir 6 CU is not that much slower vs "standard Vega 11 CU".
Only if you decide to look away the fact, acording to those numbers, that the Vega 11 on the 3400G is still faster than the Vega 7 on Renoir and that the gap to Renoir Vega 8 is half of the gap from Renoir Vega 6 to Picasso Vega 11.
That Vega 11 is overcloked. Still, at stock is about the same level has the vega 7.

Look it was already bad enoght with the downgrade, but the increase frecuency kinda makes up for it, but if they try to replace the 3400G with a Vega 6 SKU is just unaceptable, no matter anything else.
 
Last edited:

LightningZ71

Senior member
Mar 10, 2017
414
311
106
They can justify the replacement of the 3400g with the 4300g by noting that the CPU cores are considerably faster than the ones on 3400g. Yes, the gpu section is a slight regression, however, we haven't seen what its capable of with decent RAM and a decent overclock. The scores aren't night and day different at stock for GPU performance, and we know that 3200/3400g had issues with pushing RAM much past 3200-3400 speeds. Early benchmarks are showing that Renoir can push well past 4000, which can make up for a lot of the memory bandwidth limitations of the 3200/3400g. I see no reason to be disappointed at present and believe that most 4300gs will make it to a 1900mhz clock speed and support high DDR4 speeds and will be, effectively, similar to the outgoing 3400g in actual usage.

I base a lot of this on the comparison between two of my kid's laptops. I have one with a 3500u laptop and another with a 4700u one, both with similar TPU limits. While the 3500u has an additional CU over the 4700, and both have 8 threads, the 4700u is comfortably better than the 3500u in every game that we've tested so far. Now we can point at the 4700 having 8 real cores instead of 4/8 like the 3500, but, in graphics power, the raw theoretical throughput numbers aren't much different between them. However, the RAM on the 4700 is faster, and it has much better CPU performance. Both of those should be the case in comparing the 3400g to the 4300g.
 

Shivansps

Platinum Member
Sep 11, 2013
2,914
631
136
They can justify the replacement of the 3400g with the 4300g by noting that the CPU cores are considerably faster than the ones on 3400g. Yes, the gpu section is a slight regression, however, we haven't seen what its capable of with decent RAM and a decent overclock. The scores aren't night and day different at stock for GPU performance, and we know that 3200/3400g had issues with pushing RAM much past 3200-3400 speeds. Early benchmarks are showing that Renoir can push well past 4000, which can make up for a lot of the memory bandwidth limitations of the 3200/3400g. I see no reason to be disappointed at present and believe that most 4300gs will make it to a 1900mhz clock speed and support high DDR4 speeds and will be, effectively, similar to the outgoing 3400g in actual usage.

I base a lot of this on the comparison between two of my kid's laptops. I have one with a 3500u laptop and another with a 4700u one, both with similar TPU limits. While the 3500u has an additional CU over the 4700, and both have 8 threads, the 4700u is comfortably better than the 3500u in every game that we've tested so far. Now we can point at the 4700 having 8 real cores instead of 4/8 like the 3500, but, in graphics power, the raw theoretical throughput numbers aren't much different between them. However, the RAM on the 4700 is faster, and it has much better CPU performance. Both of those should be the case in comparing the 3400g to the 4300g.
I would not call a "slight regression" having a GPU that is a considerably smaller AT the same price. With that in mind any kind of GPU perf regresion is unaceptable, no matter what else.

And all this fits with the rumor that Dali replacement is a small RDNA 2 APU coming early on 2021, meaning the 4100G/4200G at $100 may be a silicion that should be used for an $60 Athlon. This is no good, this is not about making hard decisions on a power budget, this is due to having no competition on this market.
 
Last edited:

Asterox

Senior member
May 15, 2012
331
322
136
I would not call a "slight regression" having a GPU that is a considerably smaller AT the same price. With that in mind any kind of GPU perf regresion is unaceptable, no matter what else.

And all this fits with the rumor that Dali replacement is a small RDNA 2 APU coming early on 2021, meaning the 4100G/4200G at $100 may be a silicion that should be used for an $60 Athlon. This is no good, this is not about making hard decisions on a power budget, this is due to having no competition on this market.
Small RDNA 2 APU in iGPU gaming? :grinning: Compared to poor gaming big Vega, well we now how this comparison will look in iGPU gaming comparison.

Hm, in gaming small RX 5700XT/2560 SP vs Radeon VII/3846 SP + HBM2 memory or Vega 64/4096 SP+HBM2 Memory.


For example, uh small RDNA 2 will eat poor Vega for diner no doubt.

- 4/8 APU iGPU gaming, 8 CU Vega in Renoir+3600mhz system memory
- 4/8 APU iGPU gaming, small 7 CU RDNA 2 +3600mhz system memory
 

Shivansps

Platinum Member
Sep 11, 2013
2,914
631
136
Sorry, AMD does not design chips specifically for you.
Its not about the chips, this is a business decision . And if you guys gona defend this you might as well rename the forum to "political discussion" and all wear "Vote AMD" t shirts. Because that is what its going to look like.

Last year when i said that it was not good to decrese the iGPU size everyone and his mother said "wait for benchmarks". OK now we are talking about replacing a 4/8 Zen+ Vega 11@1.4Ghz with a 4/8 Zen 2 Vega 6@1.7Ghz, with a highly likely performance regression and it is still OK? SORRY, NO, IS NOT OK. PERIOD.

BTW, i still belive that the prices arent true, thats too much, even with market monopoly.
 
Last edited:

Shivansps

Platinum Member
Sep 11, 2013
2,914
631
136
Small RDNA 2 APU in iGPU gaming? :grinning: Compared to poor gaming big Vega, well we now how this comparison will look in iGPU gaming comparison.

Hm, in gaming small RX 5700XT/2560 SP vs Radeon VII/3846 SP + HBM2 memory or Vega 64/4096 SP+HBM2 Memory.


For example, uh small RDNA 2 will eat poor Vega for diner no doubt.

- 4/8 APU iGPU gaming, 8 CU Vega in Renoir+3600mhz system memory
- 4/8 APU iGPU gaming, small 7 CU RDNA 2 +3600mhz system memory
Even if is 1CU RDNA2?
 

Asterox

Senior member
May 15, 2012
331
322
136
Even if is 1CU RDNA2?
That would be stupid, or not if 1 CU RDNA 2+VCN 3.0 can run at least 8K 60hz resolution.


Renoir APU, as far i now it suports(new B550 motherboard with HDMI 2.1 suport) 8K 60hz monitor or TV.


Keep in mind, Athlon 3000G suports 4K 60hz(motherboard with HDMI 2.0 suport) and this is 50$ APU+Vega 3.

If we go by logic, AMD will continue in the same direction. 3 CU RDNA 2 GPU for low end APU models, or as expected suport for 8K 60hz.
 
Last edited:
  • Like
Reactions: Tlh97 and Elfear

Shivansps

Platinum Member
Sep 11, 2013
2,914
631
136
That would be stupid, or not if 1 CU RDNA 2+VCN 3.0 can run at least 8K 60hz resolution.


Renoir APU, as far i now it suports(new B550 motherboard with HDMI 2.1 suport) 8K 60hz monitor or TV.


Keep in mind, Athlon 3000G suports 4K 60hz(motherboard with HDMI 2.0 suport) and this is 50$ APU+Vega 3.

If we go by logic, AMD will continue in the same direction. 3 CU RDNA 2 GPU for low end APU models, or as expected suport for 8K 60hz.
Keep in mind that AMD is likely to increase core count to 4C as well, thats why i think it may have fewer than 3CUs, as long it is does perform more than Vega 3 thats enoght.

At any rate im not sure what AMD is going to do here, but everything seem to be falling in line with that rumor, if the starting Renoir ends up to be $140 they need another APU below that ASAP.
 
Last edited:

Timorous

Senior member
Oct 27, 2008
389
148
116
I would not call a "slight regression" having a GPU that is a considerably smaller AT the same price. With that in mind any kind of GPU perf regresion is unaceptable, no matter what else.
The GPU in the 5600XT is faster than the GPU in the 5700 yet due to memory bandwidth the 5700 is considerably ahead. Point being lets wait for actual in game benchmarks before we talk about a regression.
 

Shivansps

Platinum Member
Sep 11, 2013
2,914
631
136
The GPU in the 5600XT is faster than the GPU in the 5700 yet due to memory bandwidth the 5700 is considerably ahead. Point being lets wait for actual in game benchmarks before we talk about a regression.
Im willing to wait in fact i dont belive those prices to be true either, but we already know some syntetic benchmark results, and some of those are posted in the exact same page and the timespy graphics score of the 4350G is slightly over the stock 3200G with DDR4-3200 rams.


You can see that the stock 3400G with DDR4-3200 rams scores way over that 4350G

In fact, it scores slightly over the 4650G.
 

LightningZ71

Senior member
Mar 10, 2017
414
311
106
I would really like to know what the memory speed of those 4x50 chips was for that test. It is well known that the 3200/3400g got significant performance uplifts from overclocking ram (neither was specced for 3200, though could obviously support it). If we run them at spec, their scores are notably lower. Given that, if we allow the rwnoir desktop APUs to run higher clocked RAM, where are their final numbers?

I don't expect the 4300/4350g to fully surpass the 3400g on benchmarks as it's definitely at the bottom of the stack, though I do believe that actual in-game performance will be broadly comparable. What I'm more concerned about is how well the 4500 and the 4700 do.

Remember, either compare spec for spec, or allow both platforms to fully overclock. That means running the 4350 at 1900mhz+ and DDR-4xxx dram to test. Otherwise, we can just slap an LN2 pot on both processors and show anything we want.
 

Shivansps

Platinum Member
Sep 11, 2013
2,914
631
136
I would really like to know what the memory speed of those 4x50 chips was for that test. It is well known that the 3200/3400g got significant performance uplifts from overclocking ram (neither was specced for 3200, though could obviously support it). If we run them at spec, their scores are notably lower. Given that, if we allow the rwnoir desktop APUs to run higher clocked RAM, where are their final numbers?

I don't expect the 4300/4350g to fully surpass the 3400g on benchmarks as it's definitely at the bottom of the stack, though I do believe that actual in-game performance will be broadly comparable. What I'm more concerned about is how well the 4500 and the 4700 do.

Remember, either compare spec for spec, or allow both platforms to fully overclock. That means running the 4350 at 1900mhz+ and DDR-4xxx dram to test. Otherwise, we can just slap an LN2 pot on both processors and show anything we want.
No one runs IGP tests at CPU-SPEC ram, no one except AMD that did do it to compare mobile Picasso to mobile Renoir. It is really dumb to limit 3400G to 2666 and Renoir to 3200mhz just to say it is faster, it is a biased comparison.

OC results matters, yes but lets face, the people who actually do OC is a very small minority.

Anyway, we dont need to look too far, a 3200G Vega 8 needs 1600mhz to match a stock 3400G, both with DDR4-3200 rams, you can check these videos:


The Vega 6 at 1700mhz is it very very likely to be slower.

If we are talking about OCing both the 3400G and its Renoir counterpart, remember that the Vega 11 can reach 1600mhz whiout too much trouble... even 1700mhz is possible.

Were Renoir will win is with ram, with super fast DDR4 that Renoir will support (DDR4-4266), as Picasso max out at 3400mhz... but it is also super expensive.

Lets be honest here, the 4300G is a fine 3200G replacement, the 3400G is a little too much.
 

Asterox

Senior member
May 15, 2012
331
322
136
I would really like to know what the memory speed of those 4x50 chips was for that test. It is well known that the 3200/3400g got significant performance uplifts from overclocking ram (neither was specced for 3200, though could obviously support it). If we run them at spec, their scores are notably lower. Given that, if we allow the rwnoir desktop APUs to run higher clocked RAM, where are their final numbers?

I don't expect the 4300/4350g to fully surpass the 3400g on benchmarks as it's definitely at the bottom of the stack, though I do believe that actual in-game performance will be broadly comparable. What I'm more concerned about is how well the 4500 and the 4700 do.

Remember, either compare spec for spec, or allow both platforms to fully overclock. That means running the 4350 at 1900mhz+ and DDR-4xxx dram to test. Otherwise, we can just slap an LN2 pot on both processors and show anything we want.
Very soon, we will have comparison test R3 4350G 65W vs R5 3400G 65W with 3200mhz system memory.Stock setings, or no any additional overclocking as expected.
 

IntelUser2000

Elite Member
Oct 14, 2003
6,877
1,424
136
Now we can point at the 4700 having 8 real cores instead of 4/8 like the 3500, but, in graphics power, the raw theoretical throughput numbers aren't much different between them.
4700U: Vega 7 1.6GHz
3500U: Vega 8 1.2GHz

It's 17% in favor of the 4700U. Unlike in desktops where its a regression. The 3400G has 50% higher compute than 4300G. You need the top of the line 4700G to get 9% higher compute than 3400G.
 

ASK THE COMMUNITY