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

[Grim5]

That link seems to say that the 3600 is at 3.6GHz not 4.4?

It’s reporting the CPu baseclock and not boost/overclock must be a bug the 3600 will never run that slow under load

The person who uploaded the bench was the one claiming it’s a 4.4ghz all core

 

[Alan1969]

Only while in the PL2 state, and it hits about 210w.

210w is toasty. I am starting to appreciate a good cpu at stock rather than having to flog it. It costs to overclock - a good mobo, a better cooler etc and rather you could put that money towards a better cpu at stock (I am especially looking forward to the Ryzen 3000's).

 

[Potatowithearsontheside]

as i understand it, despite the fact most high end motherboards ignore the intel spec 'out of the box' the standard 95W TDP spec should be to drop to ~4.3GHz all core boost if they're respecting PL1/PL2. Obviously most of us will max these values anyway, but I was just wondering what the likelihood was that AMD tested assuming MCE so ~4.7GHz all core.

My assumption would be not! and that they'd test the 9900k @5.0 single/4.3 all core with 2666 ram, which would mean the 3900x might be a little more behind than the slides depict after most 9900k owners gain a +400MHz boost straight off the bat...

Given that it is only possible to adhere to TDP as "a" stock setting on one of the Z390 boards, I'd be fairly confident that the 9900K was allowed to go 4.7GHz all-core by AMD.

 

[chrcoluk]

p-states overclocking is the future.

A bit like on maxwell few of us were bios modding to adjust voltage bins, and now its much more widespread after voltage curve editing got implemented in afterburner.

 

[chrcoluk]

Well typically factory settings have to be very conservative for the very worst chips so there's always going to be some headroom on 99% of chips. Also they're never going to be configured for maximum performance out of the box because the TDP has to be taken into account.

True for intel for sure, but with AMD many people have tried and the gains over XFR2 have been small when achieved. (assuming using sane voltages).

Even der8auer the caseking dude said the same.

I think AMD bin their chips so the ones sold and pre configured are running close to max already.

 

[chrcoluk]

Haven’t had an AMD cpu in over 10 years.

My experience with Intels boost clocks - was that As load dropped during game the clock speed on the cpu would drop. Then under load it would shoot up again. This constant up and down of clock speed created microstuttering in the games. This is actually the reason I started o erclocking cpus m, not for performance but games run much smoother on Intel when clock speeds stay the same and don’t fluctuate

Does this happen on amd as well?

Thats not to do with overclocking but power saving features. p-states.

In windows you can adjust the p-state behaviour by adjusting the power profile, e.g. on high performance mode the cpu utilisation required to get to max turbo clocks is very low, and in games it will typically hog turbo clocks, but if you set to balanced mode you may see the clock go up and down between turbo and non turbo clocks.

Also for intel there is their new speedshift bios mode (I think rquires win10 also), and that offloads p-state management to hardware making it more responsive.

For AMD in my experience they have extremely aggressive p-states in windows, and as a result you much less likely to see clock speed fluctuation in games, I dont like it been that aggressive as its power inefficient. But its adjustable behaviour in the windows power profile so e.g. when browsing you could cap clock speed to non turbo speeds so the power is much more controlled under that type of load. However the best fix for it is p-state tuning, if I ever go AMD on my main rig it will be a priority for me to get a board that allows p-states to be tuned.

 

[Mensy]

Im pretty much a noob to all these technical things yous are all talking about and im not understanding it lol

Is a single core better than multi core for gaming ?

Im looking to build a new mini itx pc mainly for gaming and im undecided whether to wait for these ryzens or just go intel. I have a I5 4690k and im looking for a decent upgrade to last me 5+ years

 

[Mercutio]

Mutli-core is better for everything but... it requires that whatever you're running is built to work well multi-core.

Older game engines (bethesda, I'm looking at you) tend to be much more responsive to single threaded performance gains.
Newer game engines and.... most other stuff is generally _fairly_ multi-core aware.

Typically, single core figures indicate relative gaming performance. We're in the transitional period at the moment though (going into DX12 being more standard) where multi-core will start to be more important for gaming too.

 

[carbide]

This.

ahh, ok. Thought I read when the 9900k reviews dropped that MCE only pushed an all core boost of 4.7 on the 9900k. Which I thought made sense seeing as an Auto MCE function would probably push more than 1.30Vcore which would generate more heat than most were prepared for. I'm more confused as to what a 9900KS brings to the table now then, unless it's 8086K quality binning and anothe +£100!

 

[Journey]

ahh, ok. Thought I read when the 9900k reviews dropped that MCE only pushed an all core boost of 4.7 on the 9900k. Which I thought made sense seeing as an Auto MCE function would probably push more than 1.30Vcore which would generate more heat than most were prepared for. I'm more confused as to what a 9900KS brings to the table now then, unless it's 8086K quality binning and anothe +£100!

I'd be very surprised if they actually release the 9900KS, and if they do it will be a paper product only. They need to concentrate on getting the 10c/20t 1151 CPU out, with at least 4.6GHz boost speed all core at <250w, and maybe a 5.0GHz boost on 2 cores.

 

[humbug]

I'd be very surprised if they actually release the 9900KS, and if they do it will be a paper product only. They need to concentrate on getting the 10c/20t 1151 CPU out, with at least 4.6GHz boost speed all core at <250w, and maybe a 5.0GHz boost on 2 cores.

250 Watts?

The FX 9590 was 220 Watt's, in Handbreak it pulled 270 watts full system at the wall, if Intel bring this 10 core with a higher power consumption than the FX 9590 are we allowed to ridicule it? the whole internet including reviewers showed no mercy to the FX 9590 for its power draw.

 

[Journey]

250 Watts?

The FX 9590 was 220 Watt's, in Handbreak it pulled 270 watts full system at the wall, if Intel bring this 10 core with a higher power consumption than the FX 9590 are we allowed to ridicule it? the whole internet including reviewers showed no mercy to the FX 9590 for its power draw.

As we established last night the 9900K pulls 210w at 4.7GHz all-core, an extra 2-cores will add around 40w, not accounting for uncore increases.

Personally I think that 210w 4.7GHz is stupid, but half my life is spend designing low power 1U servers, and the other half doing the opposite with the likes of Threadrippers, so I am always in somewhat of a a quandary where power and efficiency are concerned.

 

[humbug]

As we established last night the 9900K pulls 210w at 4.7GHz all-core, an extra 2-cores will add around 40w, not accounting for uncore increases.

Personally I think that 210w 4.7GHz is stupid, but half my life is spend designing low power 1U servers, and the other half doing the opposite with the likes of Threadrippers, so I am always in somewhat of a a quandary where power and efficiency are concerned.

I just think it's hilarious how the tables have turned, how much faster will the 3950X be and still draw less power, not one reviewer will pickup on it let alone ridicule it for that like they did the FX 9590.

 

[4K8KW10]

There is only one saviour for intel and it's to strip the iGPU and substitute it with normal cores. An eight-core can suddenly become a sixteen-core within the same transistor and die budget.

 

[GAC]

There is only one saviour for intel and it's to strip the iGPU and substitute it with normal cores. An eight-core can suddenly become a sixteen-core within the same transistor and die budget.

and melt its way through the motherboard if clocked anywhere above 4.5ghz no doubt. you do post some utter nonsense.

 

[humbug]

Oh... that too yeah ^^^^

There is only one saviour for intel and it's to strip the iGPU and substitute it with normal cores. An eight-core can suddenly become a sixteen-core within the same transistor and die budget.

Intel's Ring Bus doesn't allow for more than 10 cores, its why the Skylake-X CPU's have an Interconnect Mesh and that isn't going to work on mainstream because the gaming IPC is around 15% lower than it is on the Ring Bus, so you will end up with GPU's that are clock for clock 10 to 15% slower in games than Zen+ let alone Zen 2.

 

[4K8KW10]

and melt its way through the motherboard if clocked anywhere above 4.5ghz no doubt. you do post some utter nonsense.

You post some utter nonsense. Most high-end motherboards have no problems to deal with 4.5GHz 16-core CPUs

Intel's Ring Bus doesn't allow for more than 10 cores, its why the Skylake-X CPU's have an Interconnect Mesh and that isn't going to work on mainstream because the gaming IPC is around 15% lower than it is on the ring Bus, so you will end up with GPU's that are clock for clock 10 to 15% slower in games than Zen+ let alone Zen 2.

Gaming IPC is as important as Cinebench score. Only relevant to those who care - I don't

 

[humbug]

You post some utter nonsense. Most high-end motherboards have no problems to deal with 4.5GHz 16-core CPUs



Gaming IPC is as important as Cinebench score. Only relevant to those who care - I don't

So you don't care about the performance of your CPU?

I agree with GAC.

 

[Jokester]

An iGPU is of no use to me, so I would rather have more cores. If Intel have issues with implementing it, then that's their problem.

 

[GAC]

You post some utter nonsense. Most high-end motherboards have no problems to deal with 4.5GHz 16-core CPUs

Gaming IPC is as important as Cinebench score. Only relevant to those who care - I don't

so you're just posting to troll now ? i get it.