AMD: Zen 2 (Ryzen/Threadripper 3000?, Epyc 8000?) Speculation, Rumours and Discussion

@tunafish is assuming AdoredTV's leaks are fake.
I don't think we have enough proof to assume they're fake or not.

We had enough proof on the day he posted them. He wasn't the first person to release them, they showed up as a pastebin dump that made the rounds around reddit/other places for a while before he picked them up. In the video where he talks about them the first time, he admitted he could not in any way verify their source, he was just forwarded the dump with no other information.

If you look at the original dump and the time it was posted, it is simply 0% credible. The issue being that it contains too much information that would not have been decided yet at that time. Such as clock speeds and SKU prices. Beyond that, it contains information that has been contradicted by AMD after it was posted, such as the "G" APUs (all the 3000-series APUs will be 12nm). Also, the clocks are nonsensical, and it got release dates wrong.

Despite all this, people are still going "but we have no proof, they might be real", even though the original "leak" was just a 0 proof pastebin dump that anyone could have manufactured in about 15 minutes. This whole thing is just another case study of just how terrible the new media "tech press" is, and how fanboy hopes makes shitty fakes make the round during the silly season before anything real gets released.

I think just for kicks I should make a similar fake for the next time around, and see if I can make a "better" one that gets more youtube screentime and tech press clicks. I think I could make one that is more credible than this one, at least...
 
I might go for the 3700X if it can clock higher than the 3900X. Not sure if I want to upgrade from X370 though.
 
I think it's great that AMD is sandbagging on the 16-core part. Coffee Lake S only goes up to 8-core, so Intel will need to wait for Ice Lake to offer over 8-cores on the low/mid-range socket.
Plus, this way they get to avoid cannibalizing most Threadripper models for a while.



We had enough proof on the day he posted them. He wasn't the first person to release them, they showed up as a pastebin dump that made the rounds around reddit/other places for a while before he picked them up. In the video where he talks about them the first time, he admitted he could not in any way verify their source, he was just forwarded the dump with no other information.

If you look at the original dump and the time it was posted, it is simply 0% credible. The issue being that it contains too much information that would not have been decided yet at that time. Such as clock speeds and SKU prices. Beyond that, it contains information that has been contradicted by AMD after it was posted, such as the "G" APUs (all the 3000-series APUs will be 12nm). Also, the clocks are nonsensical, and it got release dates wrong.

Despite all this, people are still going "but we have no proof, they might be real", even though the original "leak" was just a 0 proof pastebin dump that anyone could have manufactured in about 15 minutes. This whole thing is just another case study of just how terrible the new media "tech press" is, and how fanboy hopes makes shitty fakes make the round during the silly season before anything real gets released.
AFAIK he posted the table in a video before the pastebin dump. No info on the 3200G and 3400G is official at the moment, and AFAIK only the 3200G has been caught on camera. Only the mobile U and H series are official, and those use a different naming scheme from the desktop ones (there's no 4-core Ryzen 7 for desktop). With the new 7nm APU we might see 6-8 core models that are different from Picasso. Ryzen 5 3600G and Ryzen 7 3800G could still be a thing. They do need to face Ice Lake with 1+TFLOPs Gen11 GPUs, after all.

But now we know the info was wrong, done.

Regardless, holding up for confirmation of what comes up as true or not isn't a fanboy thing. It's just the reasonable approach in general.
I'm in no rush to discredit nor blindly believe anyone.


I think just for kicks I should make a similar fake for the next time around, and see if I can make a "better" one that gets more youtube screentime and tech press clicks. I think I could make one that is more credible than this one, at least...
If AMD includes you into a super exclusive pool of people who get review samples for rare graphics cards, I might be inclined to hold up for confirmation to stuff you say.
 
Last edited by a moderator:
32MB L3 + 4 MB L2 (for the 8 core chip)

I assume the 70MB is just 2x32MB + 2*3MB L2

Up to 8-cores get 32MB L3, and above 8-cores get 64MB L3.

Isn't this a double confirmation that the L3 is inside the chiplet, and the I/O die has no L3?

If so, how are they handling coherency between the chiplets to avoid the latency from the additional hops?
Could all Matisse I/O dies have 64MB L4 which holds only a copy of each chiplet's L3?
 
Up to 8-cores get 32MB L3, and above 8-cores get 64MB L3.

Isn't this a double confirmation that the L3 is inside the chiplet, and the I/O die has no L3?

If so, how are they handling coherency between the chiplets to avoid the latency from the additional hops?
Could all Matisse I/O dies have 64MB L4 which holds only a copy of each chiplet's L3?

My bet is they don't. Like it will always be the "weak" point of Zen. No arch is perfect...
 
Back
Top