AMD Ryzen 7 4700G is "Renoir" Desktop AM4 Processor: 8-core/16-thread with "Vega" iGPU

[Daven]

Or more possibly:

RDNA2 will be the big core GPUs named 6900(XT) and 6800(XT) and refreshed RDNA1 will get the names 6700(XT), 6600(XT), 6500(XT).

Whoops I should have used RDNA instead of Navi 1 and RDNA2 instead of Navi 2. Keep forgetting to use architecture codenames instead of product codenames.

 

[GoldenX]

I believe real improvements for graphics performance will be introduced with Ryzen 5000 APUs which are expected to be Zen 3 CPU cores and Navi/RDNA based graphic units but MX250/GT1030 levels are still fine/good for an APU and yet an improvement over previous gen APUs.

The current Vega 11 on the 3400G barely beats the 1030, I know it, i have both on my PC. AMD didn't improve IGP performance since the 2000G series, and Intel is pushing for better ones on their next products.

Compared to what? They're unmatched as far as APU are concerned, the closest anyone came to this was Intel with a Vega HBM GPU added to make one of the more intriguing products of the last decade.

Check numbers of the 3400G and the 2400G, basically zero improvement, and all the Vega bugs, like missing ReLive, bad decoding performance for 60FPS videos, and useless OpenGL drivers. THIS was the moment to switch to RDNA1. A 704SP 44TMU card can't beat a 384SP 24TMU card made in an older node.

 

[Totally]

I could definitely see next gen AMD CPUs and GPUs mimicing the strategy of current Nvidia GPU product splits (16xx and 20xx):

CPU
$50-$250
2-8 core APU
Integrated graphics
Monolithic
Zen 2/Vega

$300-$750
10-20 core CPU
Chiplet
Zen 3

GPU
$150-300
Navi 1 refresh
1408-2560 cores
DirectX 12

$350-$1000
Navi 2
2560-5120 cores
DirectX 12 Ultimate
Ray-tracing
DLSS

They already do that

They have a named(gaming/compute card) and numbered(gaming only) line of cards e.g. Fury/7### series, Vega/RX series, andd currently ????/5#00series

 

[ARF]

Check numbers of the 3400G and the 2400G, basically zero improvement, and all the Vega bugs, like missing ReLive, bad decoding performance for 60FPS videos, and useless OpenGL drivers. THIS was the moment to switch to RDNA1. A 704SP 44TMU card can't beat a 384SP 24TMU card made in an older node.

Compared to Intel, it's generations ahead in the levels of support, features and overall level.
But.... how is Intel then?

Ryzen 5 2400G to 3400G is just a move from original Zen to Zen+ with slight clock uplifts.

I hope that these APUs will get higher priority now, because they are obviously very interesting and tempting products.

 

[londiste]

Compared to Intel, it's generations ahead in the levels of support, features and overall level.

Levels of support? AMD drivers and software is usually better for dedicated GPUs but iGPUs are in a bit of a worse situation. And Intel has improved their software side considerably lately.
Features? Intel is on par or might even be slightly ahead of AMD. Direct3D, for example: https://en.wikipedia.org/wiki/Feature_levels_in_Direct3D#Support_matrix

 

[phanbuey]

In terms of performance it's far ahead but the Intel drivers are pretty good recently.

 

[SL2]

The Ryzen 5000 iGPU could end up as fast as GTX 1650 or even GTX 1660 series.

OK AdoredTV.

Dunno how you come up with things like this, honestly.

 

[Apocalypsee]

Infinity fabric is a protocol that uses pciex as an interface.

You sure? I didn't find any info regarding that, can you show where did you read that?

I didnt find those info on wikichip, or any other articles

Infinity Fabric (IF) - AMD - WikiChip

Infinity Fabric (IF) is a proprietary system interconnect architecture that facilitates data and control transmission across all linked components. This architecture is utilized by AMD's recent microarchitectures for both CPU (i.e., Zen) and graphics (e.g., Vega), and any other additional...

en.wikichip.org

 

[moproblems99]

This should become AMD's best gaming processor - a monolithic die without all the negatives of the chiplet approach.

Where in the article does it say it is not chiplet?

 

[Houd.ini]

Check numbers of the 3400G and the 2400G, basically zero improvement, and all the Vega bugs, like missing ReLive, bad decoding performance for 60FPS videos, and useless OpenGL drivers. THIS was the moment to switch to RDNA1. A 704SP 44TMU card can't beat a 384SP 24TMU card made in an older node.

"The "Renoir" iGPU features the SIMD components of "Vega," but with the display- and multimedia-engines of "Navi."

Shouldn't this help the decoding performance?

Where in the article does it say it is not chiplet?

It doesn't, but if you read a good review of Renoir you would know that it is one die/ monolithic.

 

[moproblems99]

It doesn't, but if you read a good review of Renoir you would know that it is one die/ monolithic.

Fair enough, but I generally don't read reviews of things I have no intention of buying.

 

[TheoneandonlyMrK]

You sure? I didn't find any info regarding that, can you show where did you read that?

I didnt find those info on wikichip, or any other articles

Infinity Fabric (IF) - AMD - WikiChip

Infinity Fabric (IF) is a proprietary system interconnect architecture that facilitates data and control transmission across all linked components. This architecture is utilized by AMD's recent microarchitectures for both CPU (i.e., Zen) and graphics (e.g., Vega), and any other additional...

en.wikichip.org

If you read through that what you get from it is infinity fabric is a name for a subset of data transmission and system control and monitoring protocols that work coherently across many types of Io not just one, pciex including but not limited to it, it is the basis for the high bandwidth busses.
I would wager I got this from anandtech or perhaps a slide but you're link does mention pciex as a part.

 

[watzupken]

No thanks, a 512SP Vega is too slow, unless they manage something like 2500MHz.

If you look at some of the Ryzen Renoir H series results, despite the decline in number of cores, it is actually quite a bit faster. While they did not mention any changes to it, I feel AMD may have made some tweaks to it, i.e. hybrid of RDNA and GCN, since they are going to shrink the Vega GPU from 12nm to 7nm.

The current Vega 11 on the 3400G barely beats the 1030, I know it, i have both on my PC. AMD didn't improve IGP performance since the 2000G series, and Intel is pushing for better ones on their next products.

While this is true, I think you need to consider that the Vega graphics have shared resources with the CPU and can't really shine as compared to a dedicated graphic card. The slower DDR4 ram vs GDDR5 ram doesn't help either. The point being that an integrated GPU that allows you to play modern games as high as 1080p is nothing to sneeze at.

Also it is not a good comparison when you said Intel is pushing for better IGP on their next product. Have you seen anything concrete on this product so far besides rumors? What significant improvements have Intel made on their IGP for the last 4 to 5 years besides renaming the HD graphics to UHD graphics, and improving on the decoder? If that is the case, then next gen AMD integrated graphics is going to be based on RDNA 2, so we should also expect to see improvements no?

 

[Mussels]

The iGPU performance doesn't really matter, it's simply that having an iGPU option on desktop has been the only missing part of AMD's strategy until now.

I'm sure the motherboard manufacturers are quivering in anticipation of once again being able to clog up motherboard IO panels with useless display outputs like VGA, so they don't have to put a decent amount of USB on them. Yes Biostar, i'm looking at you, kindly FECK OFF with 2000-era display connectors.



Honestly that's not a massive issue, considering that our very own lord and master W1zzard has demonstrated that 8 lanes are good enough: https://www.techpowerup.com/review/nvidia-geforce-rtx-2080-ti-pci-express-scaling/7.html - I doubt that picture will change much with the next gen of GPUs.

Hell right now due to how my mobo works, i'm running my OC'd 1080 at 8x with no issues

it's just not a problem yet, single GPU's are fine on 8x 3.0 at this stage

 

[GoldenX]

If you look at some of the Ryzen Renoir H series results, despite the decline in number of cores, it is actually quite a bit faster. While they did not mention any changes to it, I feel AMD may have made some tweaks to it, i.e. hybrid of RDNA and GCN, since they are going to shrink the Vega GPU from 12nm to 7nm.


While this is true, I think you need to consider that the Vega graphics have shared resources with the CPU and can't really shine as compared to a dedicated graphic card. The slower DDR4 ram vs GDDR5 ram doesn't help either. The point being that an integrated GPU that allows you to play modern games as high as 1080p is nothing to sneeze at.

Also it is not a good comparison when you said Intel is pushing for better IGP on their next product. Have you seen anything concrete on this product so far besides rumors? What significant improvements have Intel made on their IGP for the last 4 to 5 years besides renaming the HD graphics to UHD graphics, and improving on the decoder? If that is the case, then next gen AMD integrated graphics is going to be based on RDNA 2, so we should also expect to see improvements no?

Check the Ice Lake iGPU.

 

[londiste]

You sure? I didn't find any info regarding that, can you show where did you read that?

I didnt find those info on wikichip, or any other articles

Infinity Fabric (IF) - AMD - WikiChip

Infinity Fabric (IF) is a proprietary system interconnect architecture that facilitates data and control transmission across all linked components. This architecture is utilized by AMD's recent microarchitectures for both CPU (i.e., Zen) and graphics (e.g., Vega), and any other additional...

en.wikichip.org

This is probably the relevant part from wikichip:

The IFIS were designed so they could multiplexed with other protocols such as PCIe and SATA. They operate on TX/RX 16 differential data lanes at roughly 11 pJ/b. Those links are also aligned with the package pinout of standard PCIe lanes.

Also it is not a good comparison when you said Intel is pushing for better IGP on their next product. Have you seen anything concrete on this product so far besides rumors? What significant improvements have Intel made on their IGP for the last 4 to 5 years besides renaming the HD graphics to UHD graphics, and improving on the decoder? If that is the case, then next gen AMD integrated graphics is going to be based on RDNA 2, so we should also expect to see improvements no?

Gen11 - 32/48/64 EU in Ice Lake iGPUs - is bigger and more up to date in terms of features than Gen 9.5 - 12/24/48 EU in all the Skylake based CPUs. Gen11 as it currently is in Ice Lake is a pretty fully featured DX12 GPU. Intel EU has 8 ALUs, so G1/G3/G7 are 256/384/512 shaders vs UHD610/UHD630/IrisPlus655 at 96/192/384. For comparison, RavenRidge/Picasso main configurations are Vega8 and Vega11 at 512/704 shaders. Renoir's Vega8 is at 512 shaders but higher clock speeds.

Due to Ice Lake being what it is, Gen11 is only in 15W class CPUs right now. Intel clearly has problems with manufacturing on their 10nm but that does not mean their iGPU has not improved and does not continue to do so. The next thing to look at on their side is DG1 which should be the highest Gen12 iGPU configuration for Tiger Lake with 96 EU (768 shaders).

 

[Apocalypsee]

This is probably the relevant part from wikichip:
The IFIS were designed so they could multiplexed with other protocols such as PCIe and SATA. They operate on TX/RX 16 differential data lanes at roughly 11 pJ/b. Those links are also aligned with the package pinout of standard PCIe lanes.

It say it could multiplexed with other protocol, it didn't mention it did that internally between CPU and iGPU because if it did then the 12 PCIe lanes is wrong because it didn't account for 16 lanes for the iGPU you mentioned.

Ryzen 3 2200G - AMD - WikiChip

Ryzen 3 2200G is a 64-bit quad-core low-end performance x86 desktop microprocessor introduced by AMD in early 2018. This processor is based on AMD's Zen microarchitecture and is fabricated on a 14 nm process. The 2200G operates at a base frequency of 3.5 GHz with a TDP of 65 W and a Boost...

en.wikichip.org

 

[ARF]

OK AdoredTV.

Dunno how you come up with things like this, honestly.

Common sense, mate.

Because the smallest upcoming second generation Navi2/RDNA2 Navi 23 with 240 sq.mm and 50% higher performance per watt won't compete with GTX 1600 series.
Neither will Navi 10 or Navi 14.

See and translate this paragraph:

: : 由於 AMD 的 Zen 3 架構又將持續針對 CPU 架構進行改善，可預期除了基礎單核性能以 : 外，多核的協作延遲也將會縮短，至於 GPU 部分傳出將使用 RDNA 2 架構的 Navi 23 ， : 基於 RDNA 的 Navi 原本就較 Vega 更好，而 RDNA 2 勢必進一步提升效率，但值得注意 : 的是先前傳聞的 RDNA 2 架構應該是主打硬體光線追蹤，不過考慮到整合設計，也許 : AMD 也會如同 NVIDIA 在 Turnig 的策略，僅在高階產品支援硬體光線追蹤、較低階的 : 產品則為取消光線追蹤但採高效率 RDNA 2 架構設計，其效能或許可期待能與 NVIDIA 目 : 前提供的 GTX 16 系列抗衡；此外，雖然名稱上會是 Ryzen 5000 ，不過由於本質是第四 : 代 Ryzen APU ，故應該仍為台積電改良式 7nm 製程的產品。

Re: [情報] 傳 AMD Ryzen 5000 APU 除採 Zen 3 核心

有時候我也會覺得外面這群人也真的是很厲害 隨便抓幾個名字就可以講的跟真的一樣 : : 由於 AMD 的 Zen 3 架構又將持續針對 CPU 架構進行改善，可預期除了基礎單核性能以 : 外，多核的協作延遲也將會縮短，至於 GPU 部分傳出將使用 RDNA 2 架構的 Navi 23 ，

www.ptt.cc

 

[Flanker]

Common sense, mate.

Because the smallest upcoming second generation Navi2/RDNA2 Navi 23 with 240 sq.mm and 50% higher performance per watt won't compete with GTX 1600 series.
Neither will Navi 10 or Navi 14.

See and translate this paragraph:

Re: [情報] 傳 AMD Ryzen 5000 APU 除採 Zen 3 核心

有時候我也會覺得外面這群人也真的是很厲害 隨便抓幾個名字就可以講的跟真的一樣 : : 由於 AMD 的 Zen 3 架構又將持續針對 CPU 架構進行改善，可預期除了基礎單核性能以 : 外，多核的協作延遲也將會縮短，至於 GPU 部分傳出將使用 RDNA 2 架構的 Navi 23 ，

www.ptt.cc

I'm going to assume you don't understand Chinese. That post you linked is talking about how that claim is full of ****

 

[ARF]

Renoir with Zen 2 and Vega is already breathing behind GTX 1650.
Imagine Cezanne with Zen 3 and Navi or Navi 2. Will obliterate the old and slow GTX 1600 series

 

[londiste]

Renoir with Zen 2 and Vega is already breathing behind GTX 1650.

No, it isn't.
Stop picking the clear outlier results. Why not use the other results from the same review:

https://static.techspot.com/articles-info/2003/bench/30.png

https://static.techspot.com/articles-info/2003/bench/27.png

Spoiler

 

[ARF]

No, it isn't.

That is clearly with quite slow notebook memory. Which impacts the performance badly. Try with dual-channel DDR4-3400 or higher.

 

[londiste]

That is clearly with quite slow notebook memory. Which impacts the performance badly. Try with dual-channel DDR4-3400 or higher.

Just stop. There are couple strange results that taken out of context (like you have done several times now) suggest iGPU on 4900HS is close to GTX1650. Looking at the wider results there is a very clear and notable gap between them. I could probably find a result in the same review where for example 1065G7 has better or at least close result to 4900HS and try to use it to claim they are about equal which is obviously not the case either.

The laptop in review has dual-channel DDR4-3200.
Even assuming memory bandwidth is the big limiter, Gears 5 test shows pretty clearly that in terms of compute power GTX1650 has a lot more of it.

 

[ARF]

Just stop. There are couple strange results that taken out of context (like you have done several times now) suggest iGPU on 4900HS is close to GTX1650. Looking at the wider results there is a very clear and notable gap between them. I could probably find a result in the same review where for example 1065G7 has better or at least close result to 4900HS and try to use it to claim they are about equal which is obviously not the case either.

The laptop in review has dual-channel DDR4-3200.
Even assuming memory bandwidth is the big limiter, Gears 5 test shows pretty clearly that in terms of compute power GTX1650 has a lot more of it.

WTH? Here the topic is Cezanne, not Renoir.

You are deliberately skewing the discussion from the argument that Ryzen 5000 will be as fast as GTX 1600 series, arguing that Renoir just doesn't match which is not even correct in quite a few instances which may not matter for you but matter for me.

 

[londiste]

We have absolutely zero idea what Ryzen 5000 will be. Ryzen 4000 has not been out for long and next generation is a year or more away. I do not even remember any useful rumors flying around about the Zen3 APU. Could be anything.

4700G is Renoir.
It will come out in the (hopefully near) future.