Wednesday, September 9th 2020

AMD Announces a Red October: Zen 3 on October 8, RDNA2 on October 28

AMD CEO Dr Lisa Su on Twitter just announced AMD's next-generation Ryzen processors based on the "Zen 3" microarchitecture, and next-generation Radeon RX graphics cards based on the RDNA2 graphics architecture. AMD is promising a "red" October, with next-generation Ryzen "Zen 3" launch on October 8, and next-generation Radeon RDNA2 launch on October 28. We know for sure that AMD is referring to Ryzen and not EPYC, looking at the Socket AM4 MCM animation being used. The teaser picture for Radeon RDNA2 also hints at a new reference cooling solution with large axial fans.

Update 16:54 UTC: In a separate Tweet, AMD announced the Radeon RX 6000 series graphics cards, based on the "breakthrough" RDNA2 graphics architecture.
Sources: Dr Lisa Su (Twitter), Radeon RX (Twitter)
Add your own comment

141 Comments on AMD Announces a Red October: Zen 3 on October 8, RDNA2 on October 28

#76
tabascosauz
dragontamer5788
-snip-
That's straight up not how Matisse works...

This whole idea that AMD lags behind in the quality of its IMC is just misguided. The frequency validations and benching around the DDR4-5000 mark on both single- and dual-CCD Matisse clearly shows that isn't the case. Matisse is competitive with Comet Lake; the issue at the extreme high end is that both platforms have their own quirks and secrets regarding the relationship between stability, secondary voltages and the IMC - so as with IPC, it's hard to put a finger on who exactly is better.

Matisse's weakness is the Infinity Fabric. With IF reliably clocking far lower than the IMC at DDR4-5000 (2500MHz MCLK), there's no point to pursuing high frequency RAM for anything other than validations.

You're taking Matisse's high DRAM latency and attributing that to the memory controller. The memory controller is in the I/O die. It's Infinity Fabric thats holding back DRAM and core-core latency. Though the chiplets aren't going away, if Vermeer can expand the shared resources between 8 cores, we've already seen what that can do to core latency and performance (3300X).

Renoir is merely an iterative uplift on the Matisse IMC design, yet that same design yields far lower latency on a monolithic die and also clocks up to the moon. That latency you're seeing doesn't correlate to the IMC being shitty, champ.
Posted on Reply
#77
ShurikN
EarthDog
Im guessing not? That tweet was yesterday night at 838PM ET. Tomorrow is today? AFAIK, they blew their wad announcing the dates?

I think times got mixed up, as his tweet was 9.9. at 2AM. VC probably reported it wrong and Frank meant tomorrow as in 9.9. but from US time perspective. So the actual announcement (for the announcement lol) was supposed to be today.
Makaveli
How about adding to the discussion at hand instead of thread crapping?
And how exactly are you adding to the discussion?
Posted on Reply
#78
dinmaster
What i want to know is if these are launch dates or just pr dates like nvidias video about rtx3000. Slides with dates is underwhelming if its launch dates.
Posted on Reply
#79
EarthDog
ShurikN
I think times got mixed up, as his tweet was 9.9. at 2AM. VC probably reported it wrong and Frank meant tomorrow as in 9.9. but from US time perspective. So the actual announcement (for the announcement lol) was supposed to be today.
If you go to the actual tweet, it will show you exactly what time. Since the dude is US based, that puts it between 538 and 838 pm 9/8. Tomorrow refers to 9/9, today.
Posted on Reply
#80
BoboOOZ
EarthDog
Im guessing not? That tweet was yesterday night at 838PM ET. Tomorrow is today? AFAIK, they blew their wad announcing the dates?
I imagine AMD are gonna be super tight this time around, they don't want Nvidia to launch some Super or KO card just one day before they launch and kill their event. They might have decent marketing this time for a change, by all signs.
Posted on Reply
#81
EarthDog
BoboOOZ
I imagine AMD are gonna be super tight this time around, they don't want Nvidia to launch some Super or KO card just one day before they launch and kill their event. They might have decent marketing this time for a change, by all signs.
Thats....just about impossible (for NV to launch a super sku or something else this soon.
Posted on Reply
#82
BoboOOZ
EarthDog
Thats....just about impossible (for NV to launch a super sku or something else this soon.
From what I hear, they are an extremely reactive company, they are capable of taking decisions and implementing them in days. It doesn't even need to be a real launch, just a paper launch with the real product following one month later. I can bet they will have announcements for the middle of October if they smell something threatening.
Posted on Reply
#83
ViperXTR
YES Finally the CPU that will replace my 3770K!!!
Posted on Reply
#84
EarthDog
BoboOOZ
From what I hear, they are an extremely reactive company, they are capable of taking decisions and implementing them in days. It doesn't even need to be a real launch, just a paper launch with the real product following one month later. I can bet they will have announcements for the middle of October if they smell something threatening.
Sorry man... nothing like that...no hardware. Doesn't work like that. You cant just say we have XX when its nkt even close to being ready. Lol
Posted on Reply
#85
DemonicRyzen666
tabascosauz
That's straight up not how Matisse works...

This whole idea that AND lags behind in the quality of its IMC is just misguided. The frequency validations and benching around the DDR4-5000 mark on both single- and dual-CCD Matisse clearly shows that isn't the case. Matisse is competitive with Comet Lake; the issue at the extreme high end is that both platforms have their own quirks and secrets regarding the relationship between stability, secondary voltages and the IMC - so as with IPC, it's hard to put a finger on who exactly is better.

Matisse's weakness is the Infinity Fabric. With IF reliably clocking far lower than the IMC at DDR4-5000 (2500MHz MCLK), there's no point to pursuing high frequency RAM for anything other than validations.

You're taking Matisse's high DRAM latency and attributing that to the memory controller. The memory controller is in the I/O die. It's Infinity Fabric thats holding back DRAM and core-core latency. Though the chiplets aren't going away, if Vermeer can expand the shared resources between 8 cores, we've already seen what that can do to core latency and performance (3300X).

Renoir is merely an iterative uplift on the Matisse IMC design, yet that same design yields far lower latency on a monolithic die and also clocks up to the moon. That latency you're seeing doesn't correlate to the IMC being shitty, champ.
Actually Zen+ has more Infinity links than zen2 it has CCD to CCD for 4 chiplets. Zen2 only has a Infinity link to the IOD. (from the diagrams anyways)

More links would help and adding those Chiplet to Chiplet to skip the IOD Latency would help for Chiplets.

I really wonder why the core complexes lack infinity fabric links since CCX to CCX must also through IOD on Zen2 back to the CCX.
Posted on Reply
#86
dr0
Good news! Hopefully, RDNA2 will bring some real competition into the GPU market.

PS: not really happy about these casual references to a political regime that is actually responsible for the deaths of millions of innocent people.
Posted on Reply
#87
Turmania
I`m expecting around 200mhz speed pump and 10% gain of performance on Zen3.
I have better expectations on GPU side, but Nvdia is miles ahead.
Posted on Reply
#88
Makaveli
ShurikN

I think times got mixed up, as his tweet was 9.9. at 2AM. VC probably reported it wrong and Frank meant tomorrow as in 9.9. but from US time perspective. So the actual announcement (for the announcement lol) was supposed to be today.


And how exactly are you adding to the discussion?
i've been posting without thread crapping that's how.

And now its time to move on.
Posted on Reply
#89
Minus Infinity
Ravenas
Boosting 3950x single core performance while maintaining or increasing 3950x multicore performance will be a great accomplishment.
Why would think the two are decoupled? It's pretty much known already clocks will increase, base and boost, also ~ 15% IPC uplift, faster infinity fabric and supposedly massive gains in fpu performance. I think 20% uplift across the board is not unreasonable expectation. I am also hoping there is the rumoured 10 core variant too.
Posted on Reply
#90
MAXLD
BoboOOZ
I imagine AMD are gonna be super tight this time around, they don't want Nvidia to launch some Super or KO card just one day before they launch and kill their event. They might have decent marketing this time for a change, by all signs.
Too soon for that, imo. nVidia won't even have enough chips for this new segmentation for weeks/months, much less splitting them into more SKUs so near launch. Probably the same issue with GDDR6X. They'll surely release the Super/Ti versions when they feel the wafer yields / overall quality improve to a point of allowing for more chip "cherry picking", and that's not likely to be anytime soon.
Until then, they're prob relying on their cards being faster, + ray tracing, + all the extra nVidia perks to offer a better premium deal for that extra cost. And they have the market share, fan base, and marketing machine in full swing. They won't have much worries even if RDNA2 performance is near.

It's on AMD's hands to give them a complete proper fight now.
Posted on Reply
#91
Prima.Vera
Raendor
r u serious? Red October is called so because of the revolution of 1917. The movie was named because of the whole USSR theme, and thus relating to communist revolutio.
Actually Red October was the name of te defecting Soviet Submarine :) from the movie with the same name, which btw was a very good movie. :)
Posted on Reply
#92
Mussels
Moderprator
I want a 3080/3090 cause of the software features Nv has that i like (i've got a 580 in the house so i do get to play with their stuff too) so what i really want is AMD to make convincing slides, and Nv to drop the launch prices :P
Posted on Reply
#93
dicktracy
It's as if they are ashamed of RDNA2...
Posted on Reply
#94
Mussels
Moderprator
dicktracy
It's as if they are ashamed of RDNA2...
probably wasn't quite ready to go smashing the PR for it, but NV forced their hand despite having no material ready to go

some poor engineer is being harassed by the PR guys this very second to make convincing slides for how fast it is
Posted on Reply
#95
EarthDog
...and wondering if they are still tweaking clocks up a bit....
Posted on Reply
#96
R0H1T
EarthDog
If you go to the actual tweet, it will show you exactly what time. Since the dude is US based, that puts it between 538 and 838 pm 9/8. Tomorrow refers to 9/9, today.
Welcome to tomorrow, it's just today in a loony bin. Don't forget to fasten your seat belts, the ride will be the same as ever :toast:
Posted on Reply
#97
watzupken
AnarchoPrimitiv
I'm expecting a lot of bad faith arguments to the like of "that's too late, people will just buy Nvidia", when a majority of those people were going to buy Nvidia regardless... At least thats what historical precedent has demonstrated time and time again.... Even when AMD offers a better product and for less, Nvidia still sells more, this was especially the case in the 2000s. It's just a simple fact that thr majority of consumers don't make their purchasing decisions rationally by comparing price to performance as we would like to believe... Nvidia just has this mentia that better specs at a better price seemingly has a very difficult time overcoming. Solely for the sake of competition, I wish it was different, as nothing would be better for consumers than for AMD to grab a big chunk of market share, although I have a small feeling that RDNA2 will be the most competitive generation in years, and I think the commiserate release of the consoles, and the fact that developers will be programming for the consoles first and foremost, will help AMD tremendously this time around.
On the contrary, the fact that they are going to announce this 2 months after Ampere made me feel that the product is still not ready against Ampere. If it is ready, they could have announced it in tandem with Zen 3, instead of pushing it to the end of Oct. The benefit of announcing it earlier is to dull the competitive edge of Nvidia, but the fact that they are not doing it is troubling.

RDNA 2 is expected to be a significant improvement if I take AMD to their words, but announcing 1.5 months after Ampere means they are going to miss the upgrade train and therefore, lost sales to Nvidia. Moreover announcement is 1 part, the second part is when it will be available for sale. By end Oct, they are already in the start of the holiday season. Assuming they take another 2 to 4 weeks for the product to be widely available, they are deep in the holiday season which would cost them more sales.
Posted on Reply
#98
R0H1T
Don't worry not everyone's looking to upgrade this holiday cycle, or the new year, or the Easter. It's gonna be a long road to upgrade, for many since we aren't exactly experiencing a (economic) boom or anything!
Posted on Reply
#99
watzupken
R0H1T
Don't worry not everyone's looking to upgrade this holiday cycle, or the new year, or the Easter. It's gonna be a long road to upgrade, for many since we aren't exactly experiencing a (economic) boom or anything!
While I don't disagree with you, but those that can afford and looking to upgrade would mostly have gotten an Ampere card. So that leaves AMD with even slower sales due to economic downturn.
Posted on Reply
#100
Nkd
dragontamer5788
I doubt it. Intel's IPC is very far ahead, and memory-latency is always going to be hampered given AMD's architecture. With Intel's Icelake improving IPC (and finally moving to 10nm), I think Intel is moving faster than AMD can catch up.

AMD has the advantage of a large L3 cache. I think as long as AMD is "close enough" to Intel, the extra cores make more sense for the general purpose customer (since those extra cores mean big wins in Server, 3d modeling, video editing, and other harder compute problems).

I don't think the AMD chip needs to be the best at everything, it should be the best at what it is best at, and just "keep pace" with the Intel chips with regards to single-threaded performance.



The XBox Series X / PS5 marketing slides have been pretty nice for RDNA2 so far. The question is how much of that translates to the desktop graphics cards. Raytracing + VRS seem to be big features worth hyping (even if NVidia had them on Turing already).
Are you speaking of this on another planet? how far is that? Intel is moving far ahead of AMD? look I respect intel but I seriously have no idea what you are talking about. Yes intel is faster in gaming so far but Zen 3 is suppose to address that and improve latency even further with each core having full access to cache. let me know when they have 10nm actually going. AMD will be on 5nm by then I believe. Intel talks more about AMD then their own products. When you can have your competition advertise for you, that says alot about your product.
Posted on Reply
Add your own comment