Question AMD Rembrandt/Zen 3+ APU Speculation and Discussion

Page 46 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

izaic3

Member
Nov 19, 2019
61
96
91
Alright, so we've had some leaks so far. I don't know if any of it's been confirmed yet, as it's pretty early, but here is what I've surmised so far (massive grain of salt of course):

If if turns out to have RDNA 2 and 12 CU, I could see iGPU performance potentially almost doubling over Cezanne.

If I've made any mistakes or gotten anything wrong, please let me know. I'd also love to hear more knowledgeable people weigh in on their expectations.
 
Last edited:

itsmydamnation

Platinum Member
Feb 6, 2011
2,863
3,413
136
If only that one wasn't limited to 16GB RAM...
real men have server farms and run all vm's remote /fight me
the biggest let down for me are the speaker in it , they are so bad.... everything is so good for its price point , i dont understand why they put such bad speakers in it.

I really want to see 6800U (15W-20W) vs Alder Lake 1265U (15W). I wonder if anyone will ship the 6800U in a lower TDP state. HP did in some laptops with the 5800U, so maybe.
Lenovo has configurable operations modes , but you can now set whatever SMU params you want with Ryzen controller. So if you want to under TDP , just buy the model you want and set your own.
 

eek2121

Diamond Member
Aug 2, 2005
3,045
4,266
136
Here are the links for the above-mentioned scores.
R9 6900HX
Ryzen 6980HS

Both of them are faster than a desktop R7 5800x with Its 15758 points in CB R23, Link. They are also 38% and 50% faster than R9 5900HX, Link.


edit: These scores are just a guess and not official figures given by AMD. They are not even correct, If you read what I wrote above.

Yeah I was wondering what was going on when I clicked those links. He estimated them (incorrectly at that).

If AMD were able to squeeze 30% higher performance by moving Zen 3 to 6nm and making a few minor optimizations, we wouldn’t be getting Zen 4 this year.
 
Reactions: Tlh97

RnR_au

Golden Member
Jun 6, 2021
1,810
4,428
106
If AMD were able to squeeze 30% higher performance by moving Zen 3 to 6nm and making a few minor optimizations, we wouldn’t be getting Zen 4 this year.
How much would doubling the L1 cache give you? Zen 3+ seems to have that over Zen 3.
 
Last edited:

LightningZ71

Golden Member
Mar 10, 2017
1,659
1,942
136
With the $300+ announced retail prices that are leaking for the 6500xt, it's sad to think that AMD could have released an APU based on Rembrandt that had 16MB of Infiniti Cache on it and easily sold it at a $50-100 price premium over the 5700g and it STILL would have been the best value out there for gamers on a budget building a small system.
 

ryanjagtap

Member
Sep 25, 2021
110
132
96
With the $300+ announced retail prices that are leaking for the 6500xt, it's sad to think that AMD could have released an APU based on Rembrandt that had 16MB of Infiniti Cache on it and easily sold it at a $50-100 price premium over the 5700g and it STILL would have been the best value out there for gamers on a budget building a small system.
I think with the 5700G being a monolithic die the clocks would be lowered a considerable amount to lower heat and the performance gain wouldn't be that great than you would get by v-cache stacking.
 
Last edited by a moderator:

LightningZ71

Golden Member
Mar 10, 2017
1,659
1,942
136
I was more thinking of it being for AM5 or being used in mobile where the IfC would reduce the memory demands on the IMC and help reduce total system power.

As for clock speeds, on a process that scales to 5Ghz, I fail to see where an Infinity Cache has to be clock reduced. Now, if you are referring to the iGPU CUs themselves, you may have a point. However, RDNA2 is optimized in design to have higher clocks than Vega. Being limited to 6 WGPs is likely a bigger limitation for the 6000 series APUs if it had a 16MB IfC than clocks.

The difference? A 6000 series APU has access to more than 4GB of ram directly, so long as the PC itself has 16GB+ of system RAM and a decent bios and isn't hamstrung by having to reach across the PCIe bus to access it.
 

ryanjagtap

Member
Sep 25, 2021
110
132
96
I was more thinking of it being for AM5 or being used in mobile where the IfC would reduce the memory demands on the IMC and help reduce total system power.

As for clock speeds, on a process that scales to 5Ghz, I fail to see where an Infinity Cache has to be clock reduced. Now, if you are referring to the iGPU CUs themselves, you may have a point. However, RDNA2 is optimized in design to have higher clocks than Vega. Being limited to 6 WGPs is likely a bigger limitation for the 6000 series APUs if it had a 16MB IfC than clocks.

The difference? A 6000 series APU has access to more than 4GB of ram directly, so long as the PC itself has 16GB+ of system RAM and a decent bios and isn't hamstrung by having to reach across the PCIe bus to access it.
I think they've tried to minimize the production/design cost by not making new designs and just using the old zen 3 design + power management features on 6nm node and just changing the igpu and media engine
 

LightningZ71

Golden Member
Mar 10, 2017
1,659
1,942
136
And also changing the PCIe interface to PCIe 4.0, and increasing the USB support to include USB4/TB3. Essentially, there's nothing that wasn't touched or massively overhauled in the whole processor except for the CCX...

They minimized their required effort by leaving the CCX unchanged and overhauling everything around it. There was no additional design work needed above largely lifting most of the already completed RDNA2 segment design that was used to Navi24 (of which, it seems, all NAVI 2x parts are almost just multiples of) and using that for the iGPU in Rembrandt. Yes, it certainly makes it larger, but not drastically so. Does it consume more power? Yes, of course, it's more capable. Does the SYSTEM consume more power? Debatable as the IfC would reduce iGPU hits to DRAM. Would it be more expensive? Certainly. Would it reduce the total number of parts available to the market? Yes. There are tradeoffs all over the place.

I believe that it would have netted AMD more profit and been more useful in the bottom of the market.
 
Reactions: Joe NYC

repoman27

Senior member
Dec 17, 2018
378
535
136
Yeah I saw this too, nice to have confirmation.
TBT3 is important for me too. I am using Intel laptops and the TBT ports are useful.
The USB PD are also useful, I can charge my phone and laptop with same charger and can charge from either left or right ports.
Good thing I saw this too on the Lenovo Z13

Will switch to AMD laptop with Phoenix. I want that sweet 5nm.
My prime pre-condition for my next laptop, less than 15W CPU and I shall be able to do engineering work on it. For now I am surviving during the work from home season with desktops.

Fully implementing USB4 will make the device compatible with TBT3.

It is fairly obvious AMD do not want to stick an Intel Thunderbolt sticker on their SoC.
On top of this AMD and its partners do not need to send their device for Intel to certify.

For connecting external devices (like eGPUs and NVME drives) you need PCIe tunneling which is optional part of USB4 and which is present in in RMB.
If the laptop support USB4 and you can bet most peripheral and accessory makers will put a USB4 sticker on their eGPU or external NVME product and it will work fine.
I'm not sure there's actually confirmation of Thunderbolt interoperability here. The Verge is the only place to mention it, but that may have been an assumption.

The USB4 specification doesn't require hosts to implement either PCI Express tunneling or Thunderbolt interoperability. I cannot imagine that AMD hasn't included both in Rembrandt, but it's certainly not mandatory.

USB4 and Thunderbolt are both tunneling protocols, but they are not the same. USB4 can tunnel USB3 packets in addition to PCIe and DisplayPort, and uses slightly different signaling and channel-bonding modes. A USB4 device has to additionally support Thunderbolt Alternate Mode in order to be interoperable with Thunderbolt only devices. Thunderbolt 4, on the other hand, is USB4 with mandatory Thunderbolt Alt Mode and various other requirements as dictated by Intel for certification.

Even if AMD included everything necessary for Thunderbolt 4 in their Ryzen 6000 mobile platform, it's still entirely up to the OEMs to do the work to produce a compliant implementation and get certification if they so desire. I believe Intel is currently the only source for both Thunderbolt 4 retimers and Thunderbolt / USB4 device silicon.

Interestingly, the Rembrandt iGPU looks like it will be the first GPU to support DisplayPort 2.0 UHBR. Although it appears to be UHBR10 only, despite support for 40Gbps USB4 (20 Gbit/s signaling rate). That would still provide 38.685 Gbit/s of display bandwidth though, which is > 92% of what 48Gbps HDMI 2.1 Fixed Rate Link can do after accounting for encoding and FEC overhead.

 

uzzi38

Platinum Member
Oct 16, 2019
2,702
6,405
146
Robert's talked multiple times about USB4 supporting everything TB3 can do and wanting to buy a Rembrandt laptop which supports USB4 to use with a dock connected to a GPU and monitor, so I gonna take a wild guess and say the support is there from AMD's side.
 
Reactions: Tlh97 and scannall

DisEnchantment

Golden Member
Mar 3, 2017
1,684
6,227
136
I believe Intel is currently the only source for both Thunderbolt 4 retimers and Thunderbolt / USB4 device silicon.
They could have designed it themselves or more likely using Synopsys
The DesignWare USB4 IP supports multiple high-speed interface protocols, including USB4, DisplayPort 1.4a TX, PCI Express, and Thunderbolt 3 for efficient data transfer and simultaneous delivery of data, power, and high-resolution video through a single USB Type-C cable.

Even if AMD included everything necessary for Thunderbolt 4 in their Ryzen 6000 mobile platform, it's still entirely up to the OEMs to do the work to produce a compliant implementation and get certification if they so desire. I believe Intel is currently the only source for both Thunderbolt 4 retimers and Thunderbolt / USB4 device silicon.

TBT4 is not covered in USB4 spec. Only TBT3.
From Spec only this mentioned
2.1.5 Thunderbolt™ 3 (TBT3) Compatibility Support

Linux driver (written by Intel) detected XDomain, PCIe Tunneling, DP tunnelling for RMB so the silicon should have it.

But like mentioned, AMD very likely would not want to put Thunderbolt sticker on their chip. They can always say Fully USB4 compliant, then OEM can decide what to do with their device if they want to sent to Intel for certification to get that sticker.
But for the future not sure, what will happen if Intel moves on to TBT 5 with 80 Gbps, when AMD get enough market traction they might split the ecosystem which would suck. Apple likely would do the same.
 

DisEnchantment

Golden Member
Mar 3, 2017
1,684
6,227
136
To be fair USB4 was defined based on TBT3 when TBT4 didn't exist yet. Also my impression is that the differences between TBT3 and TBT4 are such that reaching feature parity is not unsurmountable.
Yeah, it is definitely possibly but cannot claim compliance though because I am not sure Intel will allow to put its exclusive sticker (Just going spec wise I don't really know if Intel VT-d can be certified with AMD-V, but rest seems doable in firmware and external circuitry outside of SoC)
At this point it seems Intel wants to keep TBT4 to itself, TBT3 compatibility seems more due to pressure from others or fear of fragmenting the ecosystem.
But USB4 seems enough though, so personally I don't care about TBT4. PCIe tunneling, DP tunneling, Charging/PD, 40 Gbps USB are enough for almost all use cases.
 

repoman27

Senior member
Dec 17, 2018
378
535
136
Just to clarify what I was saying, I would bet money that Rembrandt does support DisplayPort and PCI Express tunneling as well as Thunderbolt Alternate mode. However, these are all optional features for hosts per the USB4 Specification, and I have yet to see anything that constitutes confirmation from AMD that these capabilities are in fact present in Rembrandt.

They could have designed it themselves or more likely using Synopsys


TBT4 is not covered in USB4 spec. Only TBT3.
From Spec only this mentioned

Linux driver (written by Intel) detected XDomain, PCIe Tunneling, DP tunnelling for RMB so the silicon should have it.

But like mentioned, AMD very likely would not want to put Thunderbolt sticker on their chip. They can always say Fully USB4 compliant, then OEM can decide what to do with their device if they want to sent to Intel for certification to get that sticker.
But for the future not sure, what will happen if Intel moves on to TBT 5 with 80 Gbps, when AMD get enough market traction they might split the ecosystem which would suck. Apple likely would do the same.
AMD may well have followed Apple's lead and used Synopsys IP for their integrated USB4 host controller solution, but my point was slightly different.

Every Thunderbolt and USB4 peripheral device currently on the market uses discrete controllers purchased from Intel. The existing ecosystem is entirely based on Intel silicon, and interoperability with those devices is where the value proposition lies for most consumers.

Furthermore, in order for OEMs to add Thunderbolt / USB4 support to a Rembrandt platform, they need to implement several additional components (Type-C port controller, USB PD controller, and Thunderbolt / USB4 retimer). Until recently, Intel was the sole source for the retimers. However, it looks like as of last quarter, Parade has announced a suitable retimer and NXP has come out with a linear redriver.

Because additional components are required, AMD can't simply slap a "Thunderbolt 4 Certified" sticker on their APU. And another reason they can't do so is because OEM's can route those signals however they choose. For instance, it appears that ASUS didn't want to deal with retimers or Thunderbolt / USB4 certification for the ROG Zephyrus G14 (2022) GA402 and instead utilized those signals for 10Gbps USB 3.2 Type-C ports with USB PD and DisplayPort Alt Mode.

To be fair USB4 was defined based on TBT3 when TBT4 didn't exist yet. Also my impression is that the differences between TBT3 and TBT4 are such that reaching feature parity is not unsurmountable.
In order to understand the differences between Thunderbolt 4 and Thunderbolt 3, you need to understand the differences between USB4 and Thunderbolt 3, which are much better summed up in the Thunderbolt 3 Compatibility section of this document: https://www.usb.org/sites/default/files/D2T1-3 - USB4 Time Sync - Host Interface - CM - TBT3.pdf

And here's a link to all of the USB Developer Days 2019 presentations, many of which are quite good (if you're into that sort of thing): https://www.usb.org/documents?search=&category[]=64&items_per_page=50

Yeah, it is definitely possibly but cannot claim compliance though because I am not sure Intel will allow to put its exclusive sticker (Just going spec wise I don't really know if Intel VT-d can be certified with AMD-V, but rest seems doable in firmware and external circuitry outside of SoC)
At this point it seems Intel wants to keep TBT4 to itself, TBT3 compatibility seems more due to pressure from others or fear of fragmenting the ecosystem.
But USB4 seems enough though, so personally I don't care about TBT4. PCIe tunneling, DP tunneling, Charging/PD, 40 Gbps USB are enough for almost all use cases.
As you say, the certification really doesn't matter here. What does matter is working support for those items you mentioned. But I'd also add Thunderbolt Alt Mode to that list, because almost all of the devices you'll want to connect to initially will require it. That might change once USB4 device silicon takes off, but there's currently a ton of Alpine Ridge devices out there that won't work with a USB4 host without Thunderbolt Alt Mode.
 

Mopetar

Diamond Member
Jan 31, 2011
8,004
6,446
136
If the GPU market doesn't even out anytime soon, AMD could make a killing selling these as desktop APUs. 1650's are going for ~$350 right now. When the Zen 4 APUs eventually come out, we may be to the point where 60 FPS 1080 gaming can be done without needing a discrete GPU.
 
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |