AMD Zen 2 (Ryzen 3000) - *** NO COMPETITOR HINTING ***

[LePhuronn]

Yeah i'm putting it down to trolling, mind you its too good of a photoshop tho, the font is identical and exactly the same colour/shade/resolution, no cut and paste lines either, i've had a close look at it, so who knows...

Browser inspector allows you to edit the HTML directly, so you can add whatever you want and it will render identically.

In fact, I've literally done just that to illustrate

 

[humbug]

Edit: refreshed page before posting ^^^^^ in that case its looking more likely to be a troll, i'll leave my intended post intact tho for thought food.
-------------------------

Did AdoredTV say anything about Cache on Ryzen 3000? the quoted L3 is 32MB vs 16 on Ryzen 2000, the L2 has not changed at 4MB.

Is there any reason why the L3 could double in size but the L2 remain the same? doesn't look like something a troll would make up.

 

[Jokester]

32MB was mentioned in one of the AdoredTV videos as a consequence of each 8c chip having it's own 16MB L3 cache. L2 was never mentioned in it, would expect that to double as well.

 

[humbug]

32MB was mentioned in one of the AdoredTV videos as a consequence of each 8c chip having it's own 16MB L3 cache. L2 was never mentioned in it, would expect that to double as well.

Ok, that does it for me then, thanks

 

[LePhuronn]

Did AdoredTV say anything about Cache on Ryzen 3000?

The only cache talk I can recall was on the EPYC video where he mentioned 16MB L3 cache per chiplet and possibly some monstrous amount on the I/O die to hold a copy of the individual chiplet's cache to reduce the need to hop between chiplets and therefore reduce latency.

 

[humbug]

The only cache talk I can recall was on the EPYC video where he mentioned 16MB L3 cache per chiplet and possibly some monstrous amount on the I/O die to hold a copy of the individual chiplet's cache to reduce the need to hop between chiplets and therefore reduce latency.

Yeah thats similar to what Jokester is saying, i think the troll is assuming its two 8 core chiplets.

 

[LePhuronn]

I'm a bit dubious, the F6 BIOS was only released last week.

But they do have until April to get a new one out.

 

[LePhuronn]

Purely feeding the troll for a few moments, this grab is interesting in that the 3850X isn't on the list. This could be that the new Ryzens are coming a lot sooner than expected, so the CPU support page is updated now ready to rock after CES, but the 3850X is still excluded because that's not due until May and therefore won't "leak" (lol) its existence prematurely, also insinuating that AMD won't reveal the anniversary edition beast at CES.

 

[humbug]

Purely feeding the troll for a few moments, this grab is interesting in that the 3850X isn't on the list. This could be that the new Ryzens are coming a lot sooner than expected, so the CPU support page is updated now ready to rock after CES, but the 3850X is still excluded because that's not due until May and therefore won't "leak" (lol) its existence prematurely, also insinuating that AMD won't reveal the anniversary edition beast at CES.

AdoredTV said the 3850X is likely to be AMD's 40'th Anniversary Edition, very select Silicon.

AMD was founded on May First 1969, Jerry Sanders among others, he/they come mainly from Fairchild Semiconductor, that's also where Gordon Moore and Robert Noyce came from, Intel's founders, it was AMD who started designing and making Logic Processors, Intel started on Volatile Memory. (RAM)

 

[4K8KW10]

Purely feeding the troll for a few moments, this grab is interesting in that the 3850X isn't on the list. This could be that the new Ryzens are coming a lot sooner than expected, so the CPU support page is updated now ready to rock after CES, but the 3850X is still excluded because that's not due until May and therefore won't "leak" (lol) its existence prematurely, also insinuating that AMD won't reveal the anniversary edition beast at CES.

It is normal that a CPU several further months down the line isn't listed, or no?

 

[humbug]

I got my maths wrong again. 50 years, not 40. Intel's 8086 was 40 years ago, that's where i got that from, Intel celebrated their first CPU recently.

 

[Jokester]

The other thing to consider about that "leak" is that there's no other Zen2 chips on it.

 

[EsaT]

Binning. AMD are putting the best performing die in Epyc 2 (low power leakage, best frequency scaling) and progressively "worse" die in consumer parts. I don't really see anything unbelievable in the quoted figures above...

Actually server chips don't need those highest clocking dies.
What AMD would want to put into EPYCs is dies with best clocks at some fixed per die power consumption/heat output.
Power efficiency is nowadays very big deal in servers and those run only little stuff whose performance depends on highest single core/thread performance.
So many of higher clocking dies could actually end up to consumer CPUs.

 

[humbug]

The other thing to consider about that "leak" is that there's no other Zen2 chips on it.

Other CPU's have been leaked, right down to the Ryzen 3 3300.

So i don't know what to make of that,. usually AMD release the lower end Ryzen about a month or two after the mid and high range, usually its Ryzen 7, then one month Later Ryzen 5 and as much as another two months Ryzen 3 and APU's.

 

[LePhuronn]

It is normal that a CPU several further months down the line isn't listed, or no?

Well yeah, which does add a tiny bit of credence to the screen grab. If the anniversary 3850X isn't due until May, Gigabyte aren't going to list an unannounced CPU on their support list months in advance.

The other thing to consider about that "leak" is that there's no other Zen2 chips on it.

Other CPU's have been leaked, right down to the Ryzen 3 3300.

So i don't know what to make of that,. usually AMD release the lower end Ryzen about a month or two after the mid and high range, usually its Ryzen 7, then one month Later Ryzen 5 and as much as another two months Ryzen 3 and APU's.

List them closer to their official launch date, because clock/boost specifics will be kept under wraps for as long as possible. So it's likely AMD will only cover the top 3 SKUs at CES which confirms Ryzen 3000 is a thing, the top SKUs are launching imminently and then the others are rolling out over the next few months. It also means the launch of X570 can hold off until June because we won't need PCI-E 4 until Navi (and Vega II?) actually show up. Fully expecting X570 boards to be shown at CES though.

Like I said, this train of thought is just feeding the troll, but there's some logic there and a hint of what we' could see starting next week.

 

[humbug]

Not long now, 12'th? did i remember that date correctly?

 

[LePhuronn]

Also, if Ryzen 9 doesn't require X570 then support is added through a BIOS update on existing boards that the vendor believes will have sufficient power delivery to handle them.

 

[LePhuronn]

Not long now, 12'th? did i remember that date correctly?

Wednesday 9th, 5pm UK time

 

[Potatowithearsontheside]

Why do you assume TDP scales linearly? Like I said before, it's entirely possible that having all 16 cores in an AM4 package saturates the IHS to the point thermals increase higher than the other SKUs. The 8 and 12 core models have some headroom because they're not fully functioning so a linear 10W is fine, but heating up 16 cores gives zero headroom and things bump up a bit more.

I'm fully aware that won't scale linearly, except maybe at its optimum efficiency clock range. If anything, it'll go exponential after that optimum efficiency range, which simply highlights my point further; we're potentially looking at the start of that exponential curve possibly between 4.2-4.3GHz, which then doesn't bode well for clocking much higher. That is my point.
I'm concerned that the figures might suggest that, though if the 135w TDP is just a conservative figure then that curve might not start until later.
I'm simply asking the question; is there something to be concerned about?

 

[humbug]

Also, if Ryzen 9 doesn't require X570 then support is added through a BIOS update on existing boards that the vendor believes will have sufficient power delivery to handle them.

The high end X470's will, those high end Aorus, Asus and ASRock boards have 10 / 12 Phase VRMs, That's enough to power an overclocked i9 7900X

Thx for the date correction BTW.