AMD Ryzen

[Zoraptor]

Zen3 and production starting end of this year according to Ian Cutress. May well be part of a simultaneous rebrand to Zen3+ though. That and the DLSS equivalent should really improve the gaming performance of their APUs.

Didn't see it mentioned, but I presume that it will get used for GPUs as well at some stage, given them looking at MCM and having such a large cache on the 6000 series.

[Keyrock]

Kinda sucks that the just announced APUs are still using Vega graphics. I'll be looking to replace my lappy sometime in the not too distant future and I might replace it with a NUC, miniPC, whatever you want to call it, since I just output to a monitor in my truck anyway, so I don't really need a screen. An APU is conducive to a small form factor device, but Vega graphics ain't gonna cut it for 1440p. AMD is obviously capable of making an APU that can handle 1440p, since the PS5 and xXxBOxXx exist, so hopefully they release an APU with RDNA2 to the public soon.

[Zoraptor]

Funny really, Sony, MS and Samsung will all have RDNA based 'APU's before AMD itself releases one to the public, Samsung's in something as small as a phone.

I'd kind of presume the 5000 series desktop APUs are stopgaps for a Zen3 refresh- there may not be much point going RDNA2 with the integrated graphics if they're just going to choke on insufficient cache and shared DDR4 RAM. The stacked cache and (maybe) DDR5 support should move any bottleneck out a bit. They may also want to avoid people trying APU raytracing until there's more bandwidth, that performance will not exactly be great even in perfect situations.

[Keyrock]

Hell has frozen over...

 

[Zoraptor]

Be waiting a long time for them to make Radeon cards though, after what happened with XFX.

Looks like there may be a 12nm GloFo Zen3/ RDNA2 APU produced. That would certainly be a good fit for low cost Athlon desktop/ cheap laptop and embedded solution. It also makes a cheap RDNA2 based Polaris replacement more likely, though there are still potential problems there the 5500 was just plain too expensive for its performance and 7nm prices have only risen for the unreleased '6500'.

Edited July 1, 2021 by Zoraptor

[Keyrock]

tl;dw - Ryzen 6000 mobile APUs with Zen 3+ 6nm and RDNA2.

[Blodhemn]

I just went from an intel duo core to an amd 8 core. Both use 65 watts. Idles in the 30cs, ram at low 1.2 volts to keep it cool. I won't need to upgrade for another 15 years. I may be dead before then.

[Zoraptor]

Great news everyone!

The insurmountable technical difficulties that prevented 5000 series on b350/x370 (but somehow not super budget a320) motherboards have been surmounted! Well done AMD, I'm sure it has nothing to do with Intel finally being competitive again and it was just your top engineers working tirelessly on finding a solution!

(dunno wtf has happened to processor prices here but I can actually buy a 5800x for not much more than I got my 1700 all those long years ago instead of +50% too. Even a 5600x was more expensive last time I checked. Maybe it's that competition thing?)

[Bartimaeus]

Would be quite the pickle if you happened to be using one of the now unsupported CPUs and updated for some reason - you need a newer CPU in order to fix your motherboard for the old CPU, . It's likely to happen at least once somewhere in the world to someone, right?

Edited March 16, 2022 by Bartimaeus

[Keyrock]

I'll be curious to see how the 5800X3D performs once independent benchmarks are out. It's otherwise the same exact processor as the 5800X but the amount of cache is MASSIVE. Certain applications will absolutely take advantage of that and perform much better. Will games be among them? In theory, yes, but we won't know for sure until independently tested.

[Bartimaeus]

The 5800X was the worst value out of the 5000 series' desktop lineup, and it's about the same CPU at the same launch price but with seemingly slightly faster cores (...though binned at lower clock speeds than the original 5800X). Doesn't seem terribly interesting to me unless there's more to the cache stack than it seems.

[Keyrock]

That's the whole reason I'm interested. It's essentially the same processor but with waaaaaay more cache. Lisa Su is obviously going to talk up the increase in cache like it's a revolutionary advance, which is probably bollocks. I'm curious to see how much difference roughly 65 MB of extra cache makes.

Whenever we've had increases in cache before they've generally been small (e.g. 10 MB to 12 MB) and accompanied by other changes (e.g. process node, DDR generation, etc.), so it was tough to get a feel for how much difference the cache increase made. Now we have a nearly 200% increase in cache with basically no other changes. This is a unique opportunity to see how much difference cache really makes.

Edited March 16, 2022 by Keyrock

[Zoraptor]

5800x3d looks more of a proof of concept/ for early adopters than anything else. AMD is certainly all in on the big caches for everything given the design of the RDNA2+ cards.

The benchmarks may be interesting, the last big jump in cache forced one of them to be redesigned as it gave unrealistic impressions due to actually fitting inside the cache.

[ComradeYellow]

1.0% and 0.1% are where its at for gaming as it amounts for consistency (avg fps is raw speed, which is low priority because GPU's are the dominant factor in games), if benchmarks show better values for the 5800X3D I would be interested.

[ComradeYellow]

 

[Sarex]

Not really ryzen, but still...

https://www.guru3d.com/news-story/here-is-an-amd-sp5-socket-(lga6096)-for-epyc-zen4-genoa-processors-its-big.html

[Keyrock]

Impressive.

[Keyrock]

I'm waiting for an Extreme ASUS Xtreme ROG Eggstream X670E XXXtreme Edition mobo. 

[Sarex]

3 hours ago, Keyrock said:

I'm waiting for an Extreme ASUS Xtreme ROG Eggstream X670E XXXtreme Edition mobo. 

Aren't those intel?

[Zoraptor]

Oddest thing about AMD's publicity blitz is the claim of 15% performance increase- as that is what you'd expect from just the node change on an equal power basis (or 30% less power at equal performance per TMSC). They're kind of implying there are no arch IPC gains and the cache doubling doing nothing as well...

[Keyrock]

2 hours ago, Zoraptor said:

Oddest thing about AMD's publicity blitz is the claim of 15% performance increase- as that is what you'd expect from just the node change on an equal power basis (or 30% less power at equal performance per TMSC). They're kind of implying there are no arch IPC gains and the cache doubling doing nothing as well...

Who cares about IPC? They have EXTREME mobos! Granted, extreme (which is code for overpriced) mobos have existed for many years but these are MORE EXTREME! I bet they have RGBs, it doesn't get more extreme than RGBs.

[Keyrock]

Ryzen 7000 set to launch this quarter.

"Looking ahead, we're on track to launch our all-new 5nm Ryzen 7000 desktop processors and AM5 platforms later this quarter with leadership performance in gaming and content creation,"

It should be noted that this quarter ends September 30th.

[Zoraptor]

September 15 seems to be the favoured date for release on the rumour mill.

[Keyrock]

The new design is going to take some getting used to.

Edit: I would be terrified to use liquid metal on that.

Edited August 17, 2022 by Keyrock

[Keyrock]

Our lord & savior Dr. Lisa Su presenting Zen 4:

5.7 GHz

Edited August 29, 2022 by Keyrock