Discussion AMD Cezanne/Zen 3 APU Speculation and Discussion

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

dr1337

Senior member
May 25, 2020
333
565
106
Fresh leak out today, not much is known but at least 8cu's is confirmed. Probably an engineering sample, core count is unknown and clocks may not be final.

This is very interesting to me because cezanne is seemingly 8cu only, and it seems unlikely to me that AMD could squeeze any more performance out of vega. A cpu only upgrade of renoir may be lackluster compared to tigerlake's quite large GPU.

What do you guys think? Will zen 3 be a large enough improvement in APU form? Will it have full cache? Are there more than 8cus? Has AMD truly evolved vega yet again or is it more like rdna?
 

uzzi38

Platinum Member
Oct 16, 2019
2,629
5,936
146
Lucienne is apparently targeted at very specific market segments, like Chromebooks. It doesn't occupy the same space as Renoir or Cezanne.
I'm not sure of that one. 14ALN05 exists. Or was it 14ALC05? One of the two or something similar, I forget now. 14ARE05 is the Ideapad 5 with Renoir.
 
Last edited:

cortexa99

Senior member
Jul 2, 2018
319
505
136
The codename Lucienne is from a Chinese message from Weibo, and Mebiuw(it's him again) quote that message directly to Twitter. It seems that Mebiuw, this guy don't know what is Lucienne at the beginning, but he immediately quote a message which include 'Lucienne=Renoir refresh' from Weibo to Twitter.

I have to doubt Mebiuw now......seems he doesn't know anything but keep posting & quoting some doubtful message around, especially Weibo and Twitter is two totally different platform(Weibo is for chinese but twitter user is all over the world), when you keep quoting across two platform you have to consider language barrier between chinese and english. But this guy seems have no feeling about it just keep quoting and have no answer about anything.


I remember the days when before Zen1 release there's so much rumor around and I try my best to doing translate&clarify works at Overclock.net, it's a hard work.o_O

edit: there's another explain that Lucienne is belongs to Cezanne but a mobile workstation SKUs. since AMD's naming scheme is a disaster I think nothing is set in stone and be better to take a truck load of salt against every rumor.
 
Last edited:

coercitiv

Diamond Member
Jan 24, 2014
6,199
11,895
136
I'm not sure of that one. 14ALN05 exists. Or was it 14ALC05? One of the two or something similar, I forget now. 14ARE05 is the Ideapad 5 with Renoir.
14ALC05 - IdeaPad Flex 5

For others who are wondering, Lenovo uses model codes that seem to incorporate core names along screen size and manufacturer. Some existing and confirmed codes:

IIL - Intel Ice Lake
ITL - Intel Tiger Lake
ARE - AMD Renoir U
ARH - AMD Renoir H
 

uzzi38

Platinum Member
Oct 16, 2019
2,629
5,936
146
The codename Lucienne is from a Chinese message from Weibo, and Mebiuw(it's him again) quote that message directly to Twitter. It seems that Mebiuw, this guy don't know what is Lucienne at the beginning, but he immediately quote a message which include 'Lucienne=Renoir refresh' from Weibo to Twitter.

I have to doubt Mebiuw now......seems he doesn't know anything but keep posting & quoting some doubtful message around, especially Weibo and Twitter is two totally different platform(Weibo is for chinese but twitter user is all over the world), when you keep quoting across two platform you have to consider language barrier between chinese and english. But this guy seems have no feeling about it just keep quoting and have no answer about anything.


I remember the days when before Zen1 release there's so much rumor around and I try my best to doing translate&clarify works at Overclock.net, it's a hard work.o_O

edit: there's another explain that Lucienne is belongs to Cezanne but a mobile workstation SKUs. since AMD's naming scheme is a disaster I think nothing is set in stone and be better to take a truck load of salt against every rumor.
Lucienne is definitely a refresh of Renoir. They share the same CPUID.

But I'm going to say this one more time. Lucienne existing and being used in non-Chromebook devices does not mean that Cezanne is later in the year.
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,686
1,221
136
Cezanne SKUs do not cross contaminate with Lucienne SKUs. They will not share x000 series, but will be on a000-series and b000-series respectively.

Given:
Dali => 3000 series FP5
Renoir = 4000 series FP6

Thus:
Lucienne = 5000 series FP6
Cezanne = 6000 series FP7r2

There is no contamination. No Zen3 product will share its x000 series with a Zen2 product.

Van Gogh/Mero are not mainstream. They are not launching instead Rembrandt is, with FP7.

All 7nm+ products were ported to 5nm's 57CPP standard cells and 7nm+ SRAM/AMS is also re-tapeout-able on 5nm.

Q1 2020:
"But all I can say is now in N5"
-> "we have customer from -- smartphone"
-> "customer from HPC-related area"

Lucienne's BIOS blobs have it reported as Cezanne, leading to the confusion of Cezanne on FP6, when it was never FP6.

Ryzen 7 5800U = 35W <== Not Cezanne
Ryzen 7 5700U = 15W
Ryzen 5 5600U = 35W <== Not Cezanne
Ryzen 5 5500U = 15W
The TDPs are examples since there is down-TDP and up-TDP.

These upper-TDP U-models are not expected to be paired with a dGPU.
 
Last edited:

NostaSeronx

Diamond Member
Sep 18, 2011
3,686
1,221
136
I'm pretty confused.
I always thought Cezanne would be 8x Zen 3 cores, Rdna 1.5, lpddr5, Pcie 4, 7nm.

How wrong am I?
Cezanne is primarily a successor to Renoir. Meaning it is targeting the same specification window of 8-cores/8-compute units.

Zen3 = 7nm+, CDNA = 7nm+, RDNA2 = 7nm+
The socket for Cezanne is DDR4/LPDDR4X still.

-> Renoir = Zen2 + GCN CUs w/ RDNA DCN, VCN, Raster/Prim, ROPs
Lucienne being a refresh has the same specs.

-> Cezanne = Zen3 + CDNA CUs w/ RDNA2 DCN, VCN, Raster/Prim, ROPS, and TMUs
Cezanne might also have CDNA's Tensor cores.

Overall: 8x Zen3 cores, 8x CDNA1 compute units, LPDDR4X/DDR4, PCIe 4.0 might be on it$, and it might be overall be 7nm density.
$PCIe 4.0, USB 4.0 are psuedo-confirmed for Cezanne. Which it shares with Rembrandt on FP7r2.

N7 -> N7P -> N6 // CPP=57-nm & M1=57-nm*
*Given here:
40 * 1.425x = 57-nm

N7+ -> N5(N5P) -> N4 // CPP=57-nm^ & M1=38-nm*
*IanD from semiwiki:
"The only metal which changes pitch with EUV is metal 1, this reduces from 57nm (matches CPP) to 38nm (3 M1 per 2 CPP). N7+ density increase comes from better M1 routing and SDB(CPODE) cell libraries, power decrease comes from this (smaller area = shorter tracks = lower capacitance) and also ("up to...") addition of 1-fin libraries (lower capacitance and leakage) for non-critical paths."
^From dolphin: http://dolphin-ic.com/products/standard-cell/tsmc_5ff_cell.html
6-track, Ultra High Density (51nm and 57nm poly pitch)

Hence, 7nm density even if it is on a 5nm node.
 
Last edited:

A///

Diamond Member
Feb 24, 2017
4,352
3,154
136
Looks like Lucienne in the leaks is a Google exclusive APU according to Videocardz and is a Renoir refresh. Cezanne is the Zen 3 core based APUs. Videocardz editors dug up a reference to Lucienne being Renoir's illegitimate daughter.


I wonder if we'll see a 10 core laptop from AMD.
 

LightningZ71

Golden Member
Mar 10, 2017
1,627
1,898
136
It appears, and I get this only from observing Google's public actions over the last 48 hours, that they want to move the Chrome Book market up a few rungs in both capability and price. They attempted to sell high end Chrome Books in the past, but, they were rather poorly received, partly due to performance. This should make them performance competitive.
 

DrMrLordX

Lifer
Apr 27, 2000
21,629
10,841
136
They just released Picasso variants for Chromebooks.

No reason for them not to. Picasso can be pretty cheap. Heck you can get an i7 Chromebook if you want:


Point being that some CPUs are low-end enough that it's actually a bit hard to find them outside of Chromebooks. I'm reasonably certain that eh well see below.

Well, as long it means Cezanne and his replacement are on time thats fine.

At the present, nobody really knows when "on time" is. I'm still guessing March but that's reaching.

Looks like Lucienne in the leaks is a Google exclusive APU according to Videocardz and is a Renoir refresh.

Makes sense.
 

DisEnchantment

Golden Member
Mar 3, 2017
1,602
5,788
136
I thought I might add VGH here to confuse people even more

DDR5 for VGH, even before CZN?
What is this APU?
Or rather for whom is LCN and VGH?
CZN is FP6 for mainstream laptops.

drm/amdgpu: get the correct vram type for van gogh
This patch is to get the correct vram type from atombios for van gogh.
Signed-off-by: Huang Rui <ray.huang@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>

Code:
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c
index 17c010d0431f..e0e3a7e4774f 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_atomfirmware.c
@@ -149,6 +149,10 @@ static int convert_atom_mem_type_to_vram_type(struct amdgpu_device *adev,
         case LpDdr4MemType:
             vram_type = AMDGPU_VRAM_TYPE_DDR4;
             break;
+        case Ddr5MemType:
+        case LpDdr5MemType:
+            vram_type = AMDGPU_VRAM_TYPE_DDR5;
+            break;
         default:
             vram_type = AMDGPU_VRAM_TYPE_UNKNOWN;
             break;
diff --git a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
index 92fbbfb16cff..2ce79bccfc30 100644
--- a/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
+++ b/drivers/gpu/drm/amd/amdgpu/amdgpu_object.c
@@ -1063,6 +1063,7 @@ static const char *amdgpu_vram_names[] = {
     "DDR3",
     "DDR4",
     "GDDR6",
+    "DDR5"
};

/**
 

NostaSeronx

Diamond Member
Sep 18, 2011
3,686
1,221
136
I thought I might add VGH here to confuse people even more

DDR5 for VGH, even before CZN?
What is this APU?
Or rather for whom is LCN and VGH?
CZN is FP6 for mainstream laptops.
1. Van Gogh has a GDDR6 memory controller which also supports DDR5.
2. Xbox One S with DDR3 BGAs and Chuwi Aerobox with DDR3 DIMMs, same die.
3. The APU is for semi-custom not mainline solutions(think Subor Z+, Chuwi Aerobox, Atari VCS, etc).
4. While, Lucienne is a refresh and Cezanne is FP7r2 with Rembrandt.
Cezanne 8 CU (Vega)
Van Gogh 8 CU (RDNA2)
Rembrandt 12 CU (RDNA2)
CZN = 8 CU CDNA w/ RDNA2 front-end/back-end
VGH = 24 CU RDNA2, one shader engine
RMB = 24 CU RDNA2, two shader engine

Xbox Series S is Van Gogh.
vangogh.jpg

Sure, there might be an 8 CU harvested part, but it is still 24 CUs. As an example: Picasso Athlon 3150G => 4 Core/3 Vega CUs. The die still has 11 Vega CUs.

Family 17h M50h-5Fh => Family 17h M90h-9Fh => Family 19h M40h-4Fh // Same CU count
 
Last edited:

NostaSeronx

Diamond Member
Sep 18, 2011
3,686
1,221
136
Definitely sounds more like 12 CU than the 24 CU that Nosta insists to be the case.
Xbox Series S die has 24 CUs and is 197.05 mm2 in size. Technically, 12 CUs would be closer to 168 mm2. Rather, than 208 mm2 in size, thus area efficency has really dropped that generation I guess.