• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD demonstrates Ryzen 9 5900X prototype with 3D V-Cache stack chiplet design

Soldato
Joined
30 Jun 2019
Posts
7,875
You always get nit pickers don't you lol. So, if might be slightly less than 40 watts to add an additional 32MB of cache, so what?
 
Last edited:
Soldato
Joined
26 May 2014
Posts
2,952
You always get nit pickers don't you lol. So, if might be slightly less than 40 watts to add an additional 32MB of cache, so what?
The entire premise of your previous post is nonsense based on a fundemental misunderstanding of the hardware you're talking about. Hardly "nitpicking" is it?
 
Soldato
Joined
30 Jun 2019
Posts
7,875
You are welcome to try to offer a better calculation (I 'd be interested to know the answer tbh), it was just a quick and dirty one.

Note - The 5950X still uses a huge amount of power... Adding more cache is hardly going to help with that.
 
Soldato
Joined
26 Sep 2010
Posts
7,150
Location
Stoke-on-Trent
@LePhuronn - Compare the total power usage of 8 core Zen 3 CPUs, 1 with 16MB of L3 cache, one with 32MB:
https://tpucdn.com/review/amd-ryzen-7-5700g/images/power-multithread.png

So around 25w higher power usage, maybe slightly less taking into account small clock frequency differences.

So perhaps ~20w for an increase of 16MB is a fair estimate.

If we assume a 5800X with an additional 32MB of L3 cache (64MB in total) was released, I'd expect that to result in an additional 40w of power usage, putting it over 200w under full load.
I really don't think that comparison is fair or remotely accurate. 5700G is a monolithic chip, 5800X has 2 dies and an interconnect; it's the design and implementation of the Matisse package that takes 20W, not a piddly 16MiB of cache.
 
Soldato
Joined
18 Oct 2002
Posts
14,108
Location
West Midlands
If you are an example of a 'grown up', I'd rather not :)

Being unable to admit you are wrong, then trying to deflect it back at the person who calls you out, we can now all see you aren't one.

Throwing around random numbers and complete guesses with a few random slides/images doesn't help if you don't understand what you are talking about in the first place.
 
Soldato
Joined
30 Jun 2019
Posts
7,875
Maybe they could use 7nm instead of 12nm for the I/O die? Might reduce power consumption ~10 watts just by doing that. Wonder why they didn't do that for the 5000 series? Maybe for cheapness /improved yields?
 
Soldato
Joined
30 Jun 2019
Posts
7,875
@humbug - I see your point, but that's only achievable by running some of the cores at a lower clockspeed. As soon as you fix all cores at 4.5Ghz, woosh.

It's also the same 32MB of cache per CCX.
 
Last edited:
Caporegime
Joined
17 Mar 2012
Posts
47,524
Location
ARC-L1, Stanton System
@humbug - I see your point, but that's only achievable by running some of the cores at a lower clockspeed. As soon as you fix all cores at 4.5Ghz, woosh.


There is a better way of working this out.

Core power 83 watts, SoC power 9.5 watts, total 92.5 watts, total package power 107.2 watts, cores + SoC / by Package = 1.158 (15.8 watts) for the L3 and interconnect.

STNZ0dH.png
 
Soldato
Joined
30 Jun 2019
Posts
7,875
14.8 watts is not bad for the L3 cache + interconnect (I think your method of working this out is more precise :D). I suppose a 5950X with more L3 cache might be doable. Probably won't need to wait long anyway, likely 1-2 months.

I still think shrinking the I/O die to 7nm would be a good idea though, if it improves total power consumption.
 
Caporegime
Joined
17 Mar 2012
Posts
47,524
Location
ARC-L1, Stanton System
15.8 watts is not bad for the L3 cache + interconnect. I suppose a 5950X might be doable. Probably won't need to wait long anyway, iikely 1-2 months.

Yeah.
They all use about 130 watts package power out of the box, i have manages to reduce my cores power by about 22 watts with +100Mhz brining the cores power down from 105 Watts to 83 watts and boosting from 4.85Ghz to <4.95Ghz or 4.65Ghz to 4.75Ghz high stress loads.

Curve Optimiser: Negative 15
CPU boost overclock override: +100Mhz
Precision boost scaler: 4X

The socket limit is 142 watts. that gives them about 10 watts to play with and my CPU is a better than average sample, if AMD did some better binning, and 18 Months later 7nm should be yielding better quality silicon..... so i don't see why AMD couldn't fit that extra L3 on it and perhaps even get a little more out of the cores, like i have.
 
Caporegime
Joined
17 Mar 2012
Posts
47,524
Location
ARC-L1, Stanton System
One more thing, remember the FX-9590? that was also on a PGA Socket, AM3+, it had a package power of 220 Watts.

AMD are willing to throw the PGA 142 Watt limit out of the window, it literally fried cheaper not well made boards, mainly those from MSI but i ran one for a year OVERCLOCKED to 5Ghz all core on my Asus Sabertooth 990FX Rev2 and it never bothered it in the slightest.

There is no reason why AMD can't release an X590 board with higher power directives for vendors to adhere to, that would unleash the high core count Ryzen 5000 which are capable of running much higher clocks all core than they currently do with the 130 watt limits in place.
 
Last edited:
Soldato
Joined
21 Oct 2002
Posts
2,787
Location
Sunderland
Just thinking not much has been said about these bar they are coming and production starting Q4 2021 - the new 6000 series mobile is supposed to be on 6nm so thought maybe they will release these as 6000 series desktop.
6nm would be ideal as they won't have to sacrifice anything to provide the extra cache.

Probably wishful thinking though and just on 7nm
 
Back
Top Bottom