Speculation: Ryzen 3000 series

[amd6502]

no, 4 and 6.

However that would imply what seems a bit unlikely, that the IO hub has some sort of an iGPU (maybe Vega 3).

The whole lineup is bogus. Same leak as months ago with the russian retailer. Prices and not even SKU s are set that early; even if 7nm portion of the series launches as early as April, this is bogus.

 

[Topweasel]

no, 4 and 6.

However that would imply what seems a bit unlikely, that the IO hub has some sort of an iGPU (maybe Vega 3).

The whole lineup is bogus. Same leak as months ago with the russian retailer. Prices and not even SKU s are set that early; even if 7nm portion of the series launches as early as April, this is bogus.

The Rumors for them were specific 6/8.

There's also the Ryzen 3600G, an eight-core, 12-thread CPU with a 3.2GHz base clock, 4GHz boost clock and TDP of 95W. This comes with a 20 compute unit Navi 12 GPU, making it AMD's first 8-core desktop APU. It is rumored to cost $199 (around £160, AU$270) and will release third quarter of 2019.

Finally, the Ryzen 3 3300G is a six-core, 12-thread CPU with a 3.0GHz base clock, 3.8GHz boost and 65W TDP. Unlike the other Ryzen 3 processors, this one will apparently come with an integrated Navi 12 GPU with 15 compute units, making it the first ever six-core desktop APU from AMD.

https://www.techradar.com/news/amd-ryzen-3000-next-generation-16-core-processors-leaked

I am not going to say it's all bogus. But I certainly have serious doubts on the APU's.

 

[Thunder 57]

no, 4 and 6.

However that would imply what seems a bit unlikely, that the IO hub has some sort of an iGPU (maybe Vega 3).

The whole lineup is bogus. Same leak as months ago with the russian retailer. Prices and not even SKU s are set that early; even if 7nm portion of the series launches as early as April, this is bogus.

I wouldn't be surprised to see Vega 3 on the I/O chip. I think AMD wants to get to a point where all mainstream CPU's have at least basic video capability. They could still offer Vega 8/10/11 as part of the product line down the road whether it be chiplet or not and differentiate it through marketing.

 

[NostaSeronx]

imho, Matisse is purely a CPU-orienated chiplet design. It has no GPU functionality whatsoever. Renoir doesn't use chiplets whatsoever. It is a fully integrated N7 design.

Renoir should have the same CPU/GPU CU count as Picasso/Raven Ridge; 4x Zen2/11x Vega2. However, there might be more significant changes in the PSP, VCN, and IMC.

PSP => Cortex A5 ..to... Cortex A32 <-- More efficiency (FW1.x to FW2.x)
VCN => Three quality settings; Quality, Balanced, Speed ...to... x264/x265 like-options; Placebo to Ultrafast <-- Increased utility.
IMC => DDR4 ...to... DDR4/DDR5 <-- Forward-support. For top frequencies, at least 6.4 GHz(JEDEC) and at most up to 9 GHz(non-JEDEC).

CPU IP is done appears in Matisse.
GPU IP is done appears in Vega20.
Extra IP is not done?

The APU chiplet will come after the GPU chiplet. Since there will be the need for three style of common chiplets; CPU, GPU, and Fusion. No GPU common chiplet, means that there isn't a Fusion/APU common chiplet.

 

[JDG1980]

I wouldn't be surprised to see Vega 3 on the I/O chip. I think AMD wants to get to a point where all mainstream CPU's have at least basic video capability. They could still offer Vega 8/10/11 as part of the product line down the road whether it be chiplet or not and differentiate it through marketing.

AMD wouldn't actually save much die space by including only 3 CUs instead of 8 (or even 11 or 16). Look at the Polaris chips, which are on the same 14/12nm process that the Zen 2 I/O die will be. Polaris 11 has 16 CUs and is a 123mm^2 chip. Polaris 12 has only 10 CUs, but is still a 103mm^2 chip. This is because a lot of the die space on small GPUs is taken up by memory controllers and fixed-function hardware. The extra 20mm^2 of die space on Polaris 11 is enough to add 6 CUs, which implies that each CU is only about 3.3mm^2.

 

[A///]

Any idea when it'll finally release? There's nearly nothing I won't do to get my hands on one.

 

[exquisitechar]

Any idea when it'll finally release? There's nearly nothing I won't do to get my hands on one.

Mid 2019, AMD has not said anything beyond that.

 

[coffeemonster]

I'm betting the 3000G chips will use the same die as picasso. ravenridge was the same die for desktop and mobie and this new one is not on 7nm or a chiplet design. probably just a tweaked 12nm 'Ravenridge+'.

 

[moinmoin]

I'm betting the 3000G chips will use the same die as picasso. ravenridge was the same die for desktop and mobie and this new one is not on 7nm or a chiplet design. probably just a tweaked 12nm 'Ravenridge+'.

I think most people (including myself) are expecting that. The issue is for desktop Ryzen 1st and 2nd gen 4 cores were already the bottom end. 3rd gen is expected to double the max amount of cores to 16 cores. Can desktop APUs stay at meager 4 cores which will be even more lower end relative to the whole available range? I think the logical conclusion is no, and (if technically possible) a way out for AMD could be combining one CPU and GPU chiplet (allowing up to 8 cores). We'll have to see if and what AMD does there.

The other option is that AMD is not doing chiplet based APUs period, and instead 7nm based monolithic Renoir will increase core count to 8 instead. But that will keep the APU dies relatively big and low profit compared to all their other dies and products.

 

[jpiniero]

I would say Renoir is more likely to be monolithic at this point. As for what core counts Renoir would have, I would say mobile would be driving the decision. Which you could argue still means 8 cores, if only because Intel is already shipping 8 core "mobile" parts. But Intel's 15W models are still 4 cores, and will be for some time.

 

[jpiniero]

https://www.notebookcheck.net/Asus-....0.html?utm_source=dlvr.it&utm_medium=twitter

Notebookcheck's first review I think of Picasso - AMD did fix the idle so the battery life is really good for a gaming laptop.

 

[Yotsugi]

https://www.notebookcheck.net/Asus-....0.html?utm_source=dlvr.it&utm_medium=twitter

Notebookcheck's first review I think of Picasso - AMD did fix the idle so the battery life is really good for a gaming laptop.

They did?
Y E S

 

[krumme]

https://www.notebookcheck.net/Asus-....0.html?utm_source=dlvr.it&utm_medium=twitter

Notebookcheck's first review I think of Picasso - AMD did fix the idle so the battery life is really good for a gaming laptop.

Yes. Impressive results and improvements. Finally strong competition have arrived. And it got the better gpu perf.
Just got a new thinkpad t480 last week but at last there is a solid choice if battery life matters much. I would certainly have preferred the stronger amd gpu instead of the Intel solution.

 

[Markfw]

Yes. Impressive results and improvements. Finally strong competition have arrived. And it got the better gpu perf.
Just got a new thinkpad t480 last week but at last there is a solid choice if battery life matters much. I would certainly have preferred the stronger amd gpu instead of the Intel solution.

What do you think for $700 ? I have it in my shopping cart. Not a gamer, but I would like more battery life, and a little more speed. With a NVME SSD, its probably way faster than what I have.

 

[krumme]

What do you think for $700 ? I have it in my shopping cart. Not a gamer, but I would like more battery life, and a little more speed. With a NVME SSD, its probably way faster than what I have.

The t480? Excellent keyboard. Acceptable screen for office if it's the 1080 ips.
Very solid robust stuff. Unfortunately the power is from USB c (some might find it smart) but not as rock solid as the old plug.
Good 256 nvme speed.
Fairly silent and completely silent at idle.
4c kbl is a definitive upgrade vs 2c kbl. 4c is minimum for mobile forward on imo.
Cheap to upgrade 8g to two 8g sticks. Remember to turn of internal battery in bios before.

 

[Markfw]

The t480? Excellent keyboard. Acceptable screen for office if it's the 1080 ips.
Very solid robust stuff. Unfortunately the power is from USB c (some might find it smart) but not as rock solid as the old plug.
Good 256 nvme speed.
Fairly silent and completely silent at idle.
4c kbl is a definitive upgrade vs 2c kbl. 4c is minimum for mobile forward on imo.
Cheap to upgrade 8g to two 8g sticks. Remember to turn of internal battery in bios before.

No, the main unit in that link, the
ASUS FX505DY-ES51 15.6" AMD Ryzen 5 3550H (2.10 GHz) AMD Radeon RX 560X 8 GB Memory 256 GB SSD Windows 10 Home 64-bit Gaming Laptop

 

[Joe Braga]

well I would buy this laptop model but isn't available here in Brazil yet

 

[krumme]

No, the main unit in that link, the
ASUS FX505DY-ES51 15.6" AMD Ryzen 5 3550H (2.10 GHz) AMD Radeon RX 560X 8 GB Memory 256 GB SSD Windows 10 Home 64-bit Gaming Laptop

Well I wouldn't know. Squarely judging by the numbers in the test it looks like a good deal but there is a lot that can't really be measured about quality. Let us know how it is if you get it!

 

[Soulkeeper]

over 16yrs ago I bought an intel laptop with higher resolution than that .... not impressed.

 

[Topweasel]

No, the main unit in that link, the
ASUS FX505DY-ES51 15.6" AMD Ryzen 5 3550H (2.10 GHz) AMD Radeon RX 560X 8 GB Memory 256 GB SSD Windows 10 Home 64-bit Gaming Laptop

That seems like a really good laptop to game around on without being as bad as my 8 core Ryzen Asus to carry around. Just one thing worth noting is that Asus laptops aren't quite as robust as laptops from other manufacturers.

 

[Topweasel]

over 16yrs ago I bought an intel laptop with higher resolution than that .... not impressed.

Personally I think unless you are going all out getting a 17" $2500 gaming laptop with the absolute best Graphics, a "gaming" laptop shouldn't have a screen better than 1080P if for no other reason then to inflate the FPS.

Edit also at that price you are always going to be dealing with tradeoffs.

 

[Gideon]

Interesting dibits from the Ryzen DRAM Calculator creator:
https://www.overclock.net/forum/13-...1-overclocking-dram-am4-414.html#post27895416

Translation into simple language. We have:

1) New memory controller with partial error correction for nonECC memory
2) Desktop processor with two (2 CCD) chiplets on board, 32 threads maximum
3) New MBIST (Memory built-in self-test)
4) Core watchdog - is a fail/safe function used to reset a system in case the microprocessor gets lost due to address or data errors
5) XFR - at the moment I do not see anything special about it, the algorithm and limits have been updated. Scalar Controll come back with new processors.
6) Updated core control has a symmetric configuration of the active cores . In 2CCD configurations, each chiplet has its own RAM channel in order to minimize latency to memory access. 1 channel on 8 cores will be a bottleneck if you use the system in the default state.

UPD: point number 6 is questionable, perhaps there will be a special long-range interface for connecting a chiplet with IO

This is not all information which I will gladden you in the near future

The most interesting part for me was the possible multiplier to Infinity Fabric speeds, finally!

They also now allow for the clock set on the Infinity Fabric (UCLK) to select the divisor, which means we are looking at IF being clocked equal to the memory frequency at dual rate instead of single rate (like 3200MHz instead of 1600MHz), potentially. That has a lot of implications on performance if I'm reading that correctly! EXCITED!!!

I hope this multiplier has more options than 1x or 2x (like x1,5, etc), because I'm sceptical that 4+ GHz memory modules will run all that well with 4+GHz IF speeds

 

[DrMrLordX]

Hmmmm very interesting. The IF multiplier does sound exciting as well. Definitely could have used that for Zen/Zen+.

 

[BTRY B 529th FA BN]

Getting excited about Ryzen 3000! Haven't posted in a while till recently but I'm really looking forward to reading up on news etc about AMD & Ryzen

 

[Paul98]

Yeah I am looking forward to this as well, going to drop a Ryzen 3000 to replace my 1700