Discussion Beyond zen 6

[Fjodor2001]

I've suggested it before, and no one took the initiative. I say go for it.

Nothing wrong with a good old fashioned vs thread. First rule of fight club is no personal attacks. Other than that, it can be like a sportsball match up. Team supporters cheering their squad and trashing the opponent. Because you can't have versus without those elements.

Anyone fills out a butt hurt report; it'll go straight in the incinerator. Nothing worse than people that can't be good sports and have some fun.

Well, would it be a possibility to reopen this existing thread:

Discussion - Nova Lake vs Zen 6 Thread

Zen 6 zen 6 uArch 12C CCX on N2/N2P for Zen 6 Desktop/Server 32 CCX on N2/N2P Zen6C for Server Zen 6 X3D Zen 6 on N3P for Mobile Zen 6 Halo Nova Lake P Cores -> Coyote Cove P cores with 4MB Shared L2 between 2 Cores E Cores -> Arctic Wolf E cores Both Have APX/AVX 10.2 8+16/4+8/8+16 bLLC die...

forums.anandtech.com

We're going to need some place where we can compare and discuss benchmarks and differences w.r.t. architecture etc for those two CPUs. Might as well use that thread for speculation comparisons for those two CPUs too already now.

 

[DrMrLordX]

I've suggested it before, and no one took the initiative. I say go for it.

The only meaningful discussion you can have are between CPUs that actually exist, and that's mostly a settled argument. One side is winning hands-down. It tends to force discussion into meaningless "my dad can beat up your dad"-type arguments related to products that haven't even seen leaks yet. Which is why <insert future CPU here> is always the go-to since existing products don't stack up favorably.

 

[Fjodor2001]

The only meaningful discussion you can have are between CPUs that actually exist, and that's mostly a settled argument. One side is winning hands-down. It tends to force discussion into meaningless "my dad can beat up your dad"-type arguments related to products that haven't even seen leaks yet. Which is why <insert future CPU here> is always the go-to since existing products don't stack up favorably.

So basically there’s nothing to discuss even for released CPUs, and speculation about future products we should not do either. Sounds like we can close down the whole CPU forum section.

But you know that’s not true. Because it’s not as easy as one side winning hands down. There will be pros/cons of the CPUs of course. Some may win in some benchmarks, some in others. And there’s price, power consumption, perf/watt, caches, iGPU, NPU, I/O, SKU lineup, etc to consider. And they will have solved things differently w.r.t. architecture etc. Just to name a few things.

There’s a reason this forum section exists and there are hundreds of even thousands of posts in the threads discussing these technical aspects.

 

[dr1337]

So basically there’s nothing to discuss even for released CPUs, and speculation about future products we should not do either. Sounds like we can close down the whole CPU forum section.

But you know that’s not true. Because it’s not as easy as one side winning hands down. There will be pros/cons of the CPUs of course. Some may win in some benchmarks, some in others. And there’s price, power consumption, perf/watt, caches, iGPU, NPU, I/O, SKU lineup, etc to consider. And they will have solved things differently w.r.t. architecture etc. Just to name a few things.

There’s a reason this forum section exists and there are hundreds of even thousands of posts in the threads discussing these technical aspects.

thats nice and all right up until the silicon cartel rolls in with a bunch of zen 7 40% faster hype with the chief whip keeping the thread in line for 6 pages straight of back and forth binary discussion with 6 word sentences.

And remember, no other leakers but theirs can be correct! especially not MLID!

 

[DrMrLordX]

So basically there’s nothing to discuss even for released CPUs,

If there is it'd be nice if people took a shot at it, instead of talking about you-know-what-that's-inevitably-going-to-disappoint-everyone.

@dr1337

Please tell me you aren't stumping for MLID.

 

[dr1337]

Please tell me you aren't stumping for MLID.

He didn't pirate our forums with a hate thread on other leakers based on him giving them fake information, for one. Secondly, he was the only leaker correct on zen 5.

In terms of least awful leaker, for better or for worse, he is leading, and its because of the toxicity in the entire rest of the tech community.

Because its different when the 'good guys' lie to us right?

 

[marees]

He didn't pirate our forums with a hate thread on other leakers based on him giving them fake information, for one. Secondly, he was the only leaker correct on zen 5.

In terms of least awful leaker, for better or for worse, he is leading, and its because of the toxicity in the entire rest of the tech community.

Because its different when the 'good guys' lie to us right?

Best to treat each leaker as merely one source of information & do your own synthesis / build composite truth. Plus plans can always change — especially when they are hit by reality

 

[dr1337]

Best to treat each leaker as merely one source of information & do your own synthesis / build composite truth. Plus plans can always change — especially when they are hit by reality

Right so when an entire group of people with bad leaks tried to legitimize themselves by feeding another leaker bad information, what does that tell you?

I don't think its that hard to interpret what I am saying. Your response is pretty weird IMO, but maybe language barrier?

 

[DrMrLordX]

He didn't pirate our forums with a hate thread on other leakers based on him giving them fake information, for one. Secondly, he was the only leaker correct on zen 5.

In terms of least awful leaker, for better or for worse, he is leading, and its because of the toxicity in the entire rest of the tech community.

Because its different when the 'good guys' lie to us right?

Please tell me you aren't spreading the Zen5% meme again, are you? How many times do people have to post server benchmarks from last year to remind people that hey maybe the so-called toxic leakers were partially correct? Also MLID is constantly getting things wrong, and he's "pirating" the entire Internet with falsehoods.

When it comes to Zen7 and beyond, I won't be looking to MLID as a primary source for anything. He should always be treated with suspicion.

 

[dr1337]

Please tell me you aren't spreading the Zen5% meme again, are you? How many times do people have to post server benchmarks from last year to remind people that hey maybe the so-called toxic leakers were partially correct? Also MLID is constantly getting things wrong, and he's "pirating" the entire Internet with falsehoods.

When it comes to Zen7 and beyond, I won't be looking to MLID as a primary source for anything. He should always be treated with suspicion.

I literally am referencing a specific event that happened on these forums june of last year.

If you think MLID is the only prominent leaker, and that he doesn't have competition, well you've never read this thread https://forums.anandtech.com/threads/should-mlid-be-considered-a-legitimate-source.2617535/.

There is massive infighting going on and while I am not endorsing him, I am trying to point out there are people trying to influence opinions based on literally falsifying information which people actually did and this actually happened. But keep not reading my posts.

All of the leakers are lying, even the ones you think are more legit than MLID. (they are all the same, might as well be adoredTV for every username)

 

[marees]

Right so when an entire group of people with bad leaks tried to legitimize themselves by feeding another leaker bad information, what does that tell you?

I don't think its that hard to interpret what I am saying. Your response is pretty weird IMO, but maybe language barrier?

I got burned with MLID when he said RDNA 4 could launch in Q4 2024 (Oct-2024 basically). Which is why I insist that you need to observe an Epistemic Humility & maybe compose/synthesize your own version of the truth from all available pieces & sources of info

 

[DrMrLordX]

If you think MLID is the only prominent leaker, and that he doesn't have competition

. . . no. What I think is:

I got burned with MLID when he said RDNA 4 could launch in Q4 2024 (Oct-2024 basically). Which is why I insist that you need to observe an Epistemic Humility & maybe compose/synthesize your own version of the truth from all available pieces & sources of info

that stuff like this has happened too many times for anyone serious to trust MLID. That has nothing to do with any other leakers or their track records.

Per the initial post, he may have gotten the internal codenames for Zen8 and 9 correct, and if so, great.

 

[Bigos]

Your "event" was about mlid gulping fake data without any verification and making it an "exclusive leak". He even fabricated the data coming from multiple sources while there was only one.

While I believe the act was not really ethical, it also showed that you should never trust the person that would fabricate the data without batting an eye.

Stop apologizing for someone else.

 

[soresu]

Per the initial post, he may have gotten the internal codenames for Zen8 and 9 correct, and if so, great.

They certainly wouldn't be a stretch considering that aside from Zen, Zen2 and Zen5 all the codenames have been Greek myth related.

Kinda funny for AMD to be doing that, seems a bit like they are riffing on ARM, probably why the latter switched to mountain codenames in consumer markets.

 

[Jan Olšan]

Some sorts of codenames are simply going to get used in more than one company. I think lots of them had some Trinities, Morpheuses, EVOs, there was "Project Denver" at Nvidia and then an actual company (Ampere even - lol) was provisionally called Project Denver.
Greek myths are what lots of people would go for when they want a nice codename/name (Odyssey, Olympus, Athena etc)

 

[soresu]

Greek myths are what lots of people would go for when they want a nice codename/name (Odyssey, Olympus, Athena etc)

Lol true, the Vera CPU custom core is named Olympus.

 

[branch_suggestion]

https://twitter.com/x/status/2011131880817774773

We are so barack.

 

[Covfefe]

https://twitter.com/x/status/2011131880817774773

We are so barack.

The patent specifically says that using stacked cache can decrease the latency vs monolithic. I know that's been speculated in the past, but this is the first time I've seen a number put on it.

12/14 is a 14% reduction in L2 latency. Zen+ had bigger latency improvements and only improved ipc by 3%. So the latency reduction from stacked L2 and L3 would be pretty tiny, maybe 1-2%. It's still really cool though.

 

[Win2012R2]

12/14 is a 14% reduction in L2 latency. Zen+ had bigger latency improvements and only improved ipc by 3%. So the latency reduction from stacked L2 and L3 would be pretty tiny, maybe 1-2%. It's still really cool though.

They will probably use reduced latency to increase size aiming for the same latency, but with better hit ratio.

 

[adroc_thurston]

12/14 is a 14% reduction in L2 latency. Zen+ had bigger latency improvements and only improved ipc by 3%. So the latency reduction from stacked L2 and L3 would be pretty tiny, maybe 1-2%.

well it ain't about latency reduction per se, but piling up like 5 or 6 megs of private L2 at an acceptable cycle count.

 

[marees]

https://twitter.com/x/status/2011131880817774773

We are so barack.

RGT thinks this is zen 8 or later

https://twitter.com/x/status/2011497505524162759

AMD's SECRET WEAPON For Zen: Stacked L2 Cache Patent Analysis​

a new secret weapon for future Zen CPUs has been revealed! In this video, we dive into everything revealed in the newly discovered/leaked AMD patent for stacked L2 cache for Ryzen processors. There is a lot to dive into today, but this is VERY exciting for the future generations of Ryzen CPUs, and could have huge impacts on not only the hardware specs and architecture, but also unlocking new levels of performance.But just what kind of specifications and performance improvements can we expect from AMD's new Ryzen tech for stacked L2 cache? And just when will we see this new technology implemented into AMD's roadmap for Ryzen gaming CPUs?

SOURCES
https://redgamingtech.com/amds-nightmare-intels-razor-hammer-lake-serpent-lake-leaks/
https://patents.google.com/patent/US20260003794A1/en?oq=US20260003794A1
https://www.latitudeds.com/post/amd-development-of-cache-architecture-from-planar-to-3d-integration
https://globaldossier.uspto.gov/details/US/18758517/A/125173

 

[soresu]

https://twitter.com/x/status/2011131880817774773

We are so barack.

L2 3D cache is inevitable.

It's just a matter of vertical interconnect density/bandwidth.

Getting both L2 and L3 off the core die will enable a significant increase in core density per socket too - definitely in range of 1,000+ cores per 2S board.

 

[marees]

any guesses on how AMD will go about naming these processors with multiple L2 / L3 cache combos ?

 

[adroc_thurston]

any guesses on how AMD will go about naming these processors with multiple L2 / L3 cache combos ?

3D UltraCache.

 

[desrever]

GamerCache Ultra Extreme