AMD Ryzen 5000-series “Vermeer Zen 3" CPU Benchmarks Leaked (Update 2)

[rockking1379]

3 hours ago, FloRolf said:

Meanwhile my 3900X heats up like a motherf#cker with a 240 AIO. Maybe it's just my Chip but I really wouldn't want another 50% power pumping through my sad loop. 

I had that issue when I first put my 3900x into my system. Took cooler off, remounted and problem went away luckily. And I only use a 120mm AIO

[BiG StroOnZ]

16 minutes ago, AldiPrayogi said:

Wait... The hell happened with 4

 

AMD has Ryzen 4000-series on mobile already and OEM desktop too (Renoir). Thence, marketing Vermeer under "4000" too could be considered bad marketing. Instead, it's much better aligning both the next-gen parts (desktop and mobile) under the same 5000-Series nomenclature. Of course this isn't factually confirmed, but generally is the thinking behind why the change occurred.

[Bombastinator]

4 minutes ago, BiG StroOnZ said:

 

AMD has Ryzen 4000-series on mobile already and OEM desktop too (Renoir). Thence, marketing Vermeer under "4000" too could be considered bad marketing. Instead, it's much better aligning both the next-gen parts (desktop and mobile) under the same 5000-Series nomenclature. Of course this isn't factually confirmed, but generally is the thinking behind why the change occurred.

Also true, but I find the visual of a large animated number 4 bleeding it’s last in a dark alley to be more visually evocative. 

[Letgomyleghoe.]

Theoretically imagine this. Intel just released 9th gen intel 14+++++++++++ while amd introduces 7th gen ryzen. In this situation amd's marketing team could say our cpu is better because 7 8 9. 

[eeeee1]

4 hours ago, BiG StroOnZ said:

5.0GHz

we all know it wont hit this. some might, but the silicon lottery i believe is pretty stacked against you with amd

[Unixsystem]

4 minutes ago, eeeee1 said:

we all know it wont hit this. some might, but the silicon lottery i believe is pretty stacked against you with amd

As a single core boost it seems totally reasonable. Not guaranteed by any means, but 3xxx was the first gen on 7nm. With over a year of refinement on the node plus whatever architectural upgrades they've made, a couple hundred mhz doesn't seem at all out of the realm of possibility. All core boost, almost certainly not (maybe with a golden chip on a custom loop). 

[BiG StroOnZ]

9 minutes ago, eeeee1 said:

we all know it wont hit this. some might, but the silicon lottery i believe is pretty stacked against you with amd

 

I'm not going to completely agree (nor disagree) with this statement. As I'm currently in the camp where I do believe a few/couple cores will at least boost to 5GHz. Here's a good article on how that would be possible (making the 5GHz number plausible / realistic):

 

https://www.hardwaretimes.com/amd-ryzen-precision-boost-overdrive-explained-precision-boost-2-vs-precision-boost-overdrive-vs-autooc-2/

[eeeee1]

3 minutes ago, Unixsystem said:

All core boost, almost certainly not (maybe with a golden chip on a custom loop). 

 

3 minutes ago, BiG StroOnZ said:

 

I'm not going to completely agree nor disagree with this statement. As I'm currently in the camp where I do believe a few/couple cores will at least boost to 5GHz. Here's a good article on how that would be possible (making the 5GHz number plausible / realistic):

 

https://www.hardwaretimes.com/amd-ryzen-precision-boost-overdrive-explained-precision-boost-2-vs-precision-boost-overdrive-vs-autooc-2/

what i meant. but you are absolutely correct. i mean if core 0 can hit 5 ghz any gamer is happy xd

[Unixsystem]

Just now, eeeee1 said:

what i meant. but you are absolutely correct. i mean if core 0 can hit 5 ghz any gamer is happy xd

If they were to stick with 7nm for another gen or two then I'd assume all core 5ghz would be within reach. With them presumably going 5nm next gen, who knows how far that'll set clocks peed back. 

[eeeee1]

Just now, Unixsystem said:

If they were to stick with 7nm for another gen or two then I'd assume all core 5ghz would be within reach. With them presumably going 5nm next gen, who knows how far that'll set clocks peed back. 

assuming maybe that the consumer isnt using a stock cooler maybe?

[leadeater]

1 hour ago, Unixsystem said:

If they were to stick with 7nm for another gen or two then I'd assume all core 5ghz would be within reach. With them presumably going 5nm next gen, who knows how far that'll set clocks peed back. 

It may not be as bad, I think the current Zen2 products on TSMC 7nm are DUV and not EUV? I don't remember. If it is DUV and moves to EUV at 7nm or 5nm then there shouldn't be a clock hit, or an improvement if its 7nm DUV vs 7nm EUV.

 

Edit:

https://www.anandtech.com/show/15589/amd-clarifies-comments-on-7nm-7nm-for-future-products-euv-not-specified

[S w a t s o n]

I'll happily take a higher TDP to clock the shit out of 16 cores. I got 12 phases that are basically idling with my 3600x, lets go

[Bombastinator]

1 hour ago, BiG StroOnZ said:

 

I'm not going to completely agree (nor disagree) with this statement. As I'm currently in the camp where I do believe a few/couple cores will at least boost to 5GHz. Here's a good article on how that would be possible (making the 5GHz number plausible / realistic):

 

https://www.hardwaretimes.com/amd-ryzen-precision-boost-overdrive-explained-precision-boost-2-vs-precision-boost-overdrive-vs-autooc-2/

Reading your link it seems that for zen2 there are three different systems designed to push the cpu as high as it can go, at least momentarily.  I found it pretty interesting.  For one it explains why the 3950 has higher single core clocks than lower core count chips.  It’s just the standard effect of PB2.  That AMD chips with more cores always had higher clocks never really made sense to me before.
 

There is PB2, PBO, and AOC.  The way i am reading this it seems you think that for zen3, only AOC would push into the 5ghz range so it would only ever see 5ghz momentarily and only in one core.  It is possible that there may be different versions of these things for zen 3 that may do different stuff.  They did go from pb1 to pb2 after all.

[BiG StroOnZ]

31 minutes ago, Bombastinator said:

 It is possible that there may be different versions of these things for Zen 3 that may do different stuff.  They did go from pb1 to pb2 after all.

 

Probably this.  

[kakik09]

I refuse to believe in TDPs anymore. But seeing as how my 3900X is rated at 105W TDP and the 5900X's is supposedly rated higher (and I'm already struggling to tame this beast at 80C at a 4.3GHz all-core OC @1.25V) does not bode well for those of us on 280mm AIOs. 

[Bombastinator]

10 minutes ago, kakik09 said:

I refuse to believe in TDPs anymore. But seeing as how my 3900X is rated at 105W TDP and the 5900X's is supposedly rated higher (and I'm already struggling to tame this beast at 80C at a 4.3GHz all-core OC @1.25V) does not bode well for those of us on 280mm AIOs. 

I think The problem with modern chips is not inadequate cooler size. It’s inadequate heat transfer. TDP is losing accuracy I think because chips have gotten so small the materials they are made out of can’t move the heat fast enough even when presented with a cool enough place for that heat to go.

[Guest]

20 minutes ago, Bombastinator said:

I think The problem with modern chips is not inadequate cooler size. It’s inadequate heat transfer. TDP is losing accuracy I think because chips have gotten so small the materials they are made out of can’t move the heat fast enough even when presented with a cool enough place for that heat to go.

Exactly. Transistor density is a biotch.

[Bombastinator]

2 minutes ago, ShrimpBrime said:

Exactly. Transistor density is a biotch.

Something clever will have to be thought of.  I recently saw a thing about some scientists that etched channels in the silicon to use for fluid.

 

my personal thought was to build the chip in a ring and put it around a heat pipe.  Ring shaped silicon apparently has stupendous manufacturing challenges because everyone is set on things being flat.  There are advantages too though one being that silicon would be vastly cheaper to produce, and all that is necessary to move a ring is a single gear wheel.  Vastly greater simplicity for photo etching. Everything would have to be redesigned though which likely means it won’t happen.

[jagdtigger]

11 hours ago, FloRolf said:

Meanwhile my 3900X heats up like a motherf#cker with a 240 AIO. Maybe it's just my Chip but I really wouldn't want another 50% power pumping through my sad loop. 

Check Vcore, my mobo set some stupidly high value by default......

[porina]

5 hours ago, kakik09 said:

I refuse to believe in TDPs anymore. But seeing as how my 3900X is rated at 105W TDP and the 5900X's is supposedly rated higher (and I'm already struggling to tame this beast at 80C at a 4.3GHz all-core OC @1.25V) does not bode well for those of us on 280mm AIOs. 

The definition of TDP has evolved and does not represent power usage directly on either AMD or Intel CPUs. On AMD side, see my previous post about PPT, which could be the number you're looking for. 3900X is 142W PPT at stock.

 

18 minutes ago, jagdtigger said:

Check Vcore, my mobo set some stupidly high value by default......

I'd add, look at voltage under load. It is normal for higher voltages to be present while under light loads as it is needed for the higher boost clocks. Under heavy loads, clocks and voltages go down.

[jagdtigger]

14 minutes ago, porina said:

I'd add, look at voltage under load. It is normal for higher voltages to be present while under light loads as it is needed for the higher boost clocks. Under heavy loads, clocks and voltages go down.

Isnt it the opposite way?

[leadeater]

4 minutes ago, jagdtigger said:

Isnt it the opposite way?

Not on AMD, the more idle/burst loads peak the vcore voltage really high to be able to hit those peak clocks but under all core that's ~500Mhz lower clocks lower vcore is needed but more current is been drawn.

[porina]

Just now, jagdtigger said:

Isnt it the opposite way?

It does sounds counter-intuitive but makes sense if you think a bit like an overclocker.

 

Generally speaking, if a CPU runs at a lower clock, it would take a lower voltage than at higher clock. We can agree on that? Now we introduce the boost clocks and power limits. If you run a light load, the boost clock is more. You have a higher clock which needs a higher voltage. If you run a heavy load, you could run into the power limit, which reduces the clocks, and you end up with a lower voltage.

[jagdtigger]

Just now, leadeater said:

Not on AMD, the more idle burst loads peak the vcore voltage really high to be able to hit those peak clocks but under all core that's ~500Mhz lower clocks lower vcore is needed but more current is been drawn.

Good to know, but in my case the mobo set something like 1,3V+ on idle......

[leadeater]

6 minutes ago, jagdtigger said:

Good to know, but in my case the mobo set something like 1,3V+ on idle......

I also seem to remember early on at release there were some boards that were like 1.5v, that got fixed though. I personally really like the way AMD does boost clocks and voltage control, does make it much harder to talk about boost clocks and all core clocks because the actual clocks are always changing under load and are different based on the workload.

 

I haven't owned one yet but I am very much planning on purchasing a Zen3 CPU, can't wait