Speculation: Ryzen 3000 series

[ozzy702]

What's this history you keep bringing up? PCIe 1.0 launched in 2003, 2.0 launched in 2007, 3.0 launched in 2010, and 4.0 was 2017. With 5.0 slated for next year, that would be the shortest period between releases.

Stop using logic, he's always right and knows more than the rest of us. Realistically, he's correct that there really aren't many desktop use cases that need PCIE 4.0, let alone PCIE 5.0, but the 4.0 spec will be short lived at the enterprise level. PCIE 5.0 may not hit desktops for a long time due to cost but it doesn't really matter.

Props to AMD on their accomplishments over the past two years and what's to come in 2019. I really do hope that they offer a no compromise solution (productivity and gaming on the Desktop) for all of us and continue to push the pace of advancement. I fully expect Intel to answer with a vengeance once they have 7nm working, but until then AMD is going to have the advantage in every market segment.

 

[ub4ty]

What's this history you keep bringing up? PCIe 1.0 launched in 2003, 2.0 launched in 2007, 3.0 launched in 2010, and 4.0 was 2017. With 5.0 slated for next year, that would be the shortest period between releases.

The history you seem to be ignoring as well as all of the technical details I provided highlighting that this won't be heading to the desktop anytime soon. The history reflects that products hit the shelves when its decided they hit the shelves not when protocol standards are ratified. Yet, you keep rattling on and on about 5.0. I just blew apart your 5.0 timeline. Give it a rest. This thread is about Ryzen 3000 which is PCIE 4.0. I could care less what nonsense Intel is pushing around the internet about their nonexistent PCIE 5.0/DDR5 sys memory platform.

 

[ub4ty]

Stop using logic, he's always right and knows more than the rest of us. Realistically, he's correct that there really aren't many desktop use cases that need PCIE 4.0, let alone PCIE 5.0, but the 4.0 spec will be short lived at the enterprise level. PCIE 5.0 may not hit desktops for a long time due to cost but it doesn't really matter.

Props to AMD on their accomplishments over the past two years and what's to come in 2019. I really do hope that they offer a no compromise solution (productivity and gaming on the Desktop) for all of us and continue to push the pace of advancement. I fully expect Intel to answer with a vengeance once they have 7nm working, but until then AMD is going to have the advantage in every market segment.

https://www.tomshardware.com/news/pcie-4.0-5.0-pci-sig-specfication,35325.html
Article from 2017 ...
We'll See PCIe 4.0 This Year (2017), PCIe 5.0 In 2019

Yeah, essentially no. PCIE 4.0 wont be short lived. There's nothing backing that assertion.
PCIE 4.0 was 2017.. Yeah, ok. It's 2019, where is it? This is the well understood history of PCIE revs... It never comes out on time and there is nothing 'shortlived' about any of the revs. That being said, like all things, the gaps between revs have become shorter. Meanwhile, the cost are soaring which will result in targeted releases... AKA, PCIE 5.0 only being made available to enterprise for years once its made more widely availably in products/cpus in say 2022 or so.

Ryzen 3000 is hopefully equipped with PCIE 4.0. Epyc Rome was just confirmed to be PCIE 4.0 and DDR4. AMD has said nothing about PCIE 5.0/DDR5. So, it's a non-discussion as far as this thread is concerned.

AMD has done brilliantly and are the platform of choice. Intel has no platform as far as i'm concerned. I welcome the day they do and cut out the shenanigans. Any talk about Intel's future platforms is garbage until I see actual products, affordability, and a non-alphabet soup offering of CPUs, sockets, and other such convoluted nonsense. If AMD hasn't said anything about x,y,z, and its not in an official roadmap leak it's a non-story.

 

[ozzy702]

https://www.tomshardware.com/news/pcie-4.0-5.0-pci-sig-specfication,35325.html
Article from 2017 ...
We'll See PCIe 4.0 This Year (2017), PCIe 5.0 In 2019

Yeah, essentially no. PCIE 4.0 wont be short lived. There's nothing backing that assertion.
PCIE 4.0 was 2017.. Yeah, ok. It's 2019, where is it? This is the well understood history of PCIE revs... It never comes out on time and there is nothing 'shortlived' about any of the revs.

PCIE 4.0 was finalized late 2017, PCIE 5.0 is looking to be early 2019. Less than a year and a half between specs, PCIE 4.0 will be short-lived in contrast will all prior revisions. It's ok to be wrong. PCIE 4.0 being short lived outside of desktop use doesn't take anything away from ZEN2 and the accomplishments AMD has made and will continue to make.

 

[ub4ty]

PCIE 4.0 was finalized late 2017, PCIE 5.0 is looking to be early 2019. Less than a year and a half between specs, PCIE 4.0 will be short-lived in contrast will all prior revisions. It's ok to be wrong. PCIE 4.0 being short lived outside of desktop use doesn't take anything away from ZEN2 and the accomplishments AMD has made and will continue to make.

oh man.. at least you're consistent. Merry christmas

 

[Guru]

I think they are going to release a 16 core Chiplet, very similar design to their server 7nm CPU's. It's going to be the high end R7 3800x line, with the 3800 probably having threading disabled. I'm assuming the cost of these CPU's will be quite higher, but still competitive again Intel's latest 9900k offering. I'm thinking $800 for the 3800x 16 core with threading cpu, $720 for the 3800 with threading disabled and possibly slightly lower clocks.

Then the 12 core parts in the R7 3700 line, 8 core parts in the R5 3600 line, 4 core parts in the R5 3500 line. We might even see the 2500 parts be the current CCX design, but with much higher clocks due to 7nm improvements.

 

[AtenRa]

I want to see an 8C 16T Ryzen 3 with 4,5GHz Sincle Core and 4.0GHz base (95W TDP) at the same price as R5 2600X.

 

[PotatoWithEarsOnSide]

If the top end 3xxxX is a 16c 4.0GHz base, I can see it cannibalising future sales.
Where do you even go from there?
The consumer would probably buy the 6c model, keep it for a few years, and then swap it for the 16c model, and that'd keep them going for a very long time.
Short of IO changes, what would a consumer ever really need to swap to another CPU for?
Long term, I think that core proliferation is going to kill the market.

 

[moinmoin]

If the top end 3xxxX is a 16c 4.0GHz base, I can see it cannibalising future sales.
Where do you even go from there?
The consumer would probably buy the 6c model, keep it for a few years, and then swap it for the 16c model, and that'd keep them going for a very long time.
Short of IO changes, what would a consumer ever really need to swap to another CPU for?
Long term, I think that core proliferation is going to kill the market.

It will kill the market as much as the limit to 4 cores killed the market until two years ago.

 

[wilds]

If the top end 3xxxX is a 16c 4.0GHz base, I can see it cannibalising future sales.
Where do you even go from there?
The consumer would probably buy the 6c model, keep it for a few years, and then swap it for the 16c model, and that'd keep them going for a very long time.
Short of IO changes, what would a consumer ever really need to swap to another CPU for?
Long term, I think that core proliferation is going to kill the market.

So why not just get a 16c Piledriver based Opteron from the year 2012? I think we are going to be looking at several years of exciting innovations from both Intel and AMD, and core count is just going to be part of the equation.

I am so glad we are no longer stuck with 4 cores with a 5-10% performance gain each generation.

 

[ub4ty]

It will kill the market as much as the limit to 4 cores killed the market until two years ago.

If the top end 3xxxX is a 16c 4.0GHz base, I can see it cannibalising future sales.
Where do you even go from there?
The consumer would probably buy the 6c model, keep it for a few years, and then swap it for the 16c model, and that'd keep them going for a very long time.
Short of IO changes, what would a consumer ever really need to swap to another CPU for?
Long term, I think that core proliferation is going to kill the market.

All roads lead to Rome gentlemen.

 

[Kedas]

PCIE 4.0 will be short-lived

A bit wrong wording to describe it, if you mean not so many labels on the boxes, probably true, but PCIE is backward compatible so it will actually stay alive and supported as long as there is PCIE just like PCIE2 is still around because it cost more to make the higher speeds.

For AM4 to keep the price low they may do PCIE4 for the 4 lanes to the (optional) new chipset and add 4 more PCIE3 lanes there and keep the rest of the lanes on PCIE3.
It's a nice improvement and will be much cheaper than changing all to PCIE4
Every time the speed doubles the distance the signal can travel without amp in between halves. On the other hand there are enough AM4 PCIE3 boards so if you want full PCIE4 you have to be willing to pay the extra price.

 

[Insert_Nickname]

For AM4 to keep the price low they may do PCIE4 for the 4 lanes to the (optional) new chipset and add 4 more PCIE3 lanes there and keep the rest of the lanes on PCIE3.
It's a nice improvement and will be much cheaper than changing all to PCIE4

Perhaps make one of the root complexes PCIe 4.0 capable? You'd get a PCIe 4.0 link to the chipset, and a dedicated PCIe 4.0 x4 for a single SSD. GPUs are still PCIe 3.0, so it wont be until the next generation after Turing/Navi(?) we'd get a 4.0 capable GPU.

But I'd settle for a PCIe 3.0 upgrade to the chipset provided lanes, and a few more lanes from it. 8 GPP lanes is a little on the low side. A PCIe 4.0 link to the CPU would just be a nice bonus.

 

[PotatoWithEarsOnSide]

Navi will be PCIe4, hence AMD saying that it makes sense for a PCIe4 GPU having access to a PCIe4 CPU, fuelling the speculation that Ryzen 3xxx would also be PCIe4.

 

[Topweasel]

Navi will be PCIe4, hence AMD saying that it makes sense for a PCIe4 GPU having access to a PCIe4 CPU, fuelling the speculation that Ryzen 3xxx would also be PCIe4.

It would be a good opportunity. But it's hard to say Epyc including PCIe 4.0 is CPU arch agnostic. It's entirely possible that if it's not using the same IO die or an IO die at all. That Ryzen 3k could still be PCIe 3. I think the determining factor is the release window for Navi. If its mid, to a little late 2019. Then Ryzen 3k probably will support PCIe 4. If its late 2019 to early 2020. Then it's probably saved for Zen 3. Though it would make sense for AMD to support PCIe 4.0 for 2 or more generations before swapping out to 5.0.

With all that in mind. I think Ryzen 3k will support PCIe 4.0. People wanting to keep their boards can still upgrade. But it gives people a reason to look to upgrade their boards if for no other reason then getting 2 full speed 4x PCIe 3.0 M.2 slots or an extra 8x PCIe 3.0 slot (would now support 3 8x slots) a good reason to do so.

 

[maddie]

Navi will be PCIe4, hence AMD saying that it makes sense for a PCIe4 GPU having access to a PCIe4 CPU, fuelling the speculation that Ryzen 3xxx would also be PCIe4.

Vega20 is already PCIe 4.0.

 

[PotatoWithEarsOnSide]

In that case, Ryzen 3xxx supporting PCIe4 seems quite likely.

 

[Topweasel]

Vega20 is already PCIe 4.0.

Yeah but it's also not meant for consumers and is server or workstation only. So what it uses and what Epyc/TR uses has little bearing on Ryzen 3k.

 

[jpiniero]

Yeah but it's also not meant for consumers and is server or workstation only. So what it uses and what Epyc/TR uses has little bearing on Ryzen 3k.

Seems like it'd be far too much engineering effort to do anything other than the 1/4 IO die.

 

[Topweasel]

Agreed. My thoughts are those without an IO chip. If Ryzen 3k is using an IO chip even if it's not the one Epyc is using, then surely it will have PCIe 4.0 support if for no other reason then they can use that for more than one Gen.

 

[maddie]

Seems like it'd be far too much engineering effort to do anything other than the 1/4 IO die.

Yep. Otherwise its doing extra work to stay behind.

 

[Kenmitch]

https://wccftech.com/amd-launching-ryzen-3000-series-cpus-apus-radeon-gpu-ces/amp/

Guess we'll find out soon enough if his souces are reliable.

 

[Glo.]

https://twitter.com/x/status/1077041301210591232

 

[chrisjames61]

I want to see an 8C 16T Ryzen 3 with 4,5GHz Sincle Core and 4.0GHz base (95W TDP) at the same price as R5 2600X.

Don't you want to see AMD make money? They aren't a charity. Why would they sell an next generation eight core for what they were selling a last generation six core? A Ryzen 3 eight core should go for what a 2700X currently goes for at a minimum and probably more.

 

[maddie]

Don't you want to see AMD make money? They aren't a charity. Why would they sell an next generation eight core for what they were selling a last generation six core? A Ryzen 3 eight core should go for what a 2700X currently goes for at a minimum and probably more.

Margin is just one part of the equation.