Speculation: Ryzen 4000 series/Zen 3

[uzzi38]

We don't know about the headroom, and Zen 3 offers one unlike Zen 2. The noteworthy part it's a new stepping. OPN is the same so they are likely not distinct products in the market.

That's what I'm thinking more or less. Seems like a silent update with potentially some firmware changes.

Or weird other theory, perhaps it's related to X570S? A new IOD with changes to drop power consumption somewhat?

 

[maddie]

That's what I'm thinking more or less. Seems like a silent update with potentially some firmware changes.

Or weird other theory, perhaps it's related to X570S? A new IOD with changes to drop power consumption somewhat?

Speaking about that. How much power does the IOD in Ryzen desktop really use? I can't remember seeing that info anywhere.

 

[coercitiv]

Speaking about that. How much power does the IOD in Ryzen desktop really use? I can't remember seeing that info anywhere.

The difference between core and un-core package and core-only power as reported by sensors can go as high as 16-21W. A big portion of this is the IOD.



From the review:

If we look directly at the Ryzen 9 5950X for chip wide power consumption over per-core loading, we get this following graph. Here we are reporting two of the values that we have access to on the chip, which the chip estimates as part of its turbo detection and action algorithms: total package power (for the whole chip), and the power solely used by the sum of cores, which includes the L3 cache. The difference between the two covers the IO die as well as any chiplet-to-chiplet communications, PCIe, CPU-to-chipset, and DRAM controller consumption.

 

[maddie]

The difference between core and un-core power as reported by sensors can go as high as 16-21W. A big portion of this is the IOD.

View attachment 44536

From the review:

Yeah. After posing the question, I realized it was indirectly given already.

I got ~ 15W for the single CPU chiplet SKUs, 5600X & 5800X.
35W for the 5900X.
21W for the 5950X.

It seems that AMD is binning the IOD also and not just the CPU chiplets. Should have thought of that, as it appears obvious in retrospect.

 

[fleshconsumed]

nm

 

[yuri69]

Back in the days of A64, Phenom II, or C2D/Q having multiple revisions of the same SKU was quite common. Some were better than others and thus preferred by end-users.

 

[Justinus]

Yeah. After posing the question, I realized it was indirectly given already.

I got ~ 15W for the single CPU chiplet SKUs, 5600X & 5800X.
35W for the 5900X.
21W for the 5950X.

It seems that AMD is binning the IOD also and not just the CPU chiplets. Should have thought of that, as it appears obvious in retrospect.

If they were binning the IOD for the 5950x I would expect 5950x samples would more easily hit higher FCLK frequencies, but that does not appear to be the case. Considering the IOD power is a function of the various voltages, memory configuration, and the FCLK, I would be surprised to see such a large power difference explained by the IOD bin vs. the motherboard's automatic voltage selection and these other factors.

Are you sure there was not a memory configuration, memory frequency, FCLK, voltage, or motherboard difference between the tested 5900x and 5950x?

 

[maddie]

If they were binning the IOD for the 5950x I would expect 5950x samples would more easily hit higher FCLK frequencies, but that does not appear to be the case. Considering the IOD power is a function of the various voltages, memory configuration, and the FCLK, I would be surprised to see such a large power difference explained by the IOD bin vs. the motherboard's automatic voltage selection and these other factors.

Are you sure there was not a memory configuration, memory frequency, FCLK, voltage, or motherboard difference between the tested 5900x and 5950x?

Got them from this https://www.anandtech.com/show/1621...ive-review-5950x-5900x-5800x-and-5700x-tested

Identical configuration except for CPUs.

 

[jpiniero]

That 6 core has the exact same specifications as the 5600X. Same base, same boost.

Not so sure about these being an XT style refresh.

Maybe the XT models will be limited to just the 12 and 16 core models. Mainly because they are also going to increase prices.

 

[scineram]

I don’t think any new SKU is coming. At least I don’t want to see a cosmetic XT update. Just focus on RMB supply.

 

[LightningZ71]

It does sort of make sense though? Perhaps they are moving to 12lp+ for the IOD? If they shave off about 5-10 watts of package power, that should allow a marginally higher all core boost while staying in total package power specs. Since the IOD is supposedly the same between the CPU package and the 570 chipset, then an x570s that doesn't require active cooling would be a natural follow. Means cheaper board construction costs for partners and a higher MSRP for the top end SKUs. It's not enough of a difference to affect the lower core count parts that have lower binned CCDs in a notable way.

 

[Kedas]

It's a new stepping according to that rumor, turbo +0.1Mhz?
You don't need a new stepping for that, you get that with binning.
and if it's a new stepping then all new dies will be made with the new one, they will not keep producing B0 when they have a B2 version.
Some bug fixes sure that I would understand.

 

[Ajay]

Back in the days of A64, Phenom II, or C2D/Q having multiple revisions of the same SKU was quite common. Some were better than others and thus preferred by end-users.

Intel Q6600 G0 - 4.4GHz Max (ran it at 4.0 most of the time).

 

[Det0x]

All current 5900 and 5950x's have one good CCD and one average/bad CCD

Give me two good CCD's on a hypothetical 5950XT and i would upgrade. (no problem selling my old 5950x)
Next on my wish list is 6nm IO die that can run higher fclk speeds

 

[Hans Gruber]

Didn't we have this same conversation when the 3600XT was introduced last generation? A nice boost in clock speeds in current gen CPU's that didn't translate into value or significant performance boost. If the Zen3 refresh is the same thing. It would only have success because of the short supply of Zen 3 CPU's. The real boost in performance comes from a new architecture like Zen 4 and DDR5 memory chips.

People question how these performance boost generation over generation are possible. Because AMD made redactedCPU's for so long and Intel had no competition. Now that AMD has a pipeline of CPU architecture. We see impressive gains between CPU generations.





No profanity allowed in the technical forums.


esquared
Anandtech Forum Director

 

[lobz]

Maybe the XT models will be limited to just the 12 and 16 core models. Mainly because they are also going to increase prices.

Increasing prices would be the most idiotic step AMD's taken in a while. Which unfortunately makes it totally possible

 

[LP-ZX100C]

Increasing prices would be the most idiotic step AMD's taken in a while. Which unfortunately makes it totally possible

In your mind ,AMD needs to sell a Ferrari LaFerrari performance - like cpu while having a price of Toyota Camry , right?

 

[uzzi38]

AMD confirms B2 stepping for Ryzen 5000 series brings no functionality or performance improvements - VideoCardz.com

AMD Ryzen 5000 “Vermeer” on B2 stepping is only to improve production and availability Yesterday it was reported that AMD is developing a new stepping of its Vermeer processors. Meanwhile, the Polish website Benchmark.pl has reached out to AMD for a comment on this matter. The site received an...

videocardz.com

It's just a new stepping for yield purposes most likely.

 

[cortexa99]

Maybe B2 with some better OC headroom? I watched Zen2 OC headroom increasing with just later production date but still old stepping, for example R5-3600 which were produced between 2020-2H to 2021-1H could do 4.5Ghz with 1.3xv and is not required to win silicon lottery.

 

[uzzi38]

Maybe B2 with some better OC headroom? I watched Zen2 OC headroom increasing with just later production date but still old stepping, for example R5-3600 which were produced between 2020-2H to 2021-1H could do 4.5Ghz with 1.3xv and is not required to win silicon lottery.

"No performance improvements" indicates AMD hasn't gone out of their way to change anything specifically for higher frequencies in the design or the packaging. But seeing as it comes a while after the B0 stepping, I guess on average B2 stepping dies may clock better just down to node maturity improving.

 

[DrMrLordX]

In your mind ,AMD needs to sell a Ferrari LaFerrari performance - like cpu while having a price of Toyota Camry , right?

Please don't use statements like this as a justification for infinite price increases. AMD already jacked prices up on Vermeer by +$50 at every price point vs. Matisse. And that's just MSRP. People have been paying gouge prices above that for about six months, which mercifully has come to an end.

Yes, Vermeer is excellent. No, Intel offers no real competition. The opportunity is there for AMD to look at the gouge prices people were willing to pay for Vermeer and test the waters with Raphael. As someone who likes to actually be able to buy AMD products and wants other people on this forum to be able to buy AMD products, I do not welcome the thought of AMD using their current position to squeeze customers any further. Their volume alone has driven significant improvements in revenue. There's little need for them to continue pushing higher price points, even though we all know they can probably get away with it.

 

[B-Riz]

Intel Q6600 G0 - 4.4GHz Max (ran it at 4.0 most of the time).

Yes! I bought a used Q6600 years ago, not sure if it was G0 or not, but man, this stuff was fun to see happening.

Way to biff it in the past few years Intel

Intel Core 2 Q6600 G0 Stepping: Cheap Quad Core Just Got Better

www.anandtech.com

 

[B-Riz]

Please don't use statements like this as a justification for infinite price increases. AMD already jacked prices up on Vermeer by +$50 at every price point vs. Matisse. And that's just MSRP. People have been paying gouge prices above that for about six months, which mercifully has come to an end.

Yes, Vermeer is excellent. No, Intel offers no real competition. The opportunity is there for AMD to look at the gouge prices people were willing to pay for Vermeer and test the waters with Raphael. As someone who likes to actually be able to buy AMD products and wants other people on this forum to be able to buy AMD products, I do not welcome the thought of AMD using their current position to squeeze customers any further. Their volume alone has driven significant improvements in revenue. There's little need for them to continue pushing higher price points, even though we all know they can probably get away with it.

AMD had the "fastest" gaming CPU's at launch, a halo product stack, how was the price increase the end of the world? Those were X products, not regular 65W parts.

Intel has the 11900K at $549 right now, that is their halo product. Whether it is worth that does not matter, it is the halo CPU for them right now.

And now the 5800X is down to $421.05 shipped and sold by Amazon. So yes, AMD knew the 5800X was over priced at $450.

They are still selling Zen2 desktop as the more budget option and it is still a very good CPU, and, more Zen3 is coming. It is now clear they had huge OEM obligations to fill for APU and mobile, which, is very smart in the grand scheme of things.

But, Intel is still there for a budget build, you are not forced to buy AMD.

 

[LP-ZX100C]

The opportunity is there for AMD to look at the gouge prices people were willing to pay for Vermeer and test the waters with Raphael.

And ,yet, i remember during Bulldozer era when Intel jacked up their prices to ridiculous levels , (1000$ for 6c and 1700$ for 8c) ,i guess then was pretty much okay , then.

Please don't use statements like this as a justification for infinite price increases.

Oh ,i'll use every time i can .(when performance is there)

There's little need for them to continue pushing higher price points, even though we all know they can probably get away with it.

I guess, everyone wants to see a weak AMD ,unable to compete with its counterparts and everyone knows the consequences.

 

[DrMrLordX]

AMD had the "fastest" gaming CPU's at launch, a halo product stack, how was the price increase the end of the world?

It wasn't, it's just annoying. And the worst part is that it was impossible to get Vermeer at MSRP for a long time, except maybe a 5800x.

I don't really care if it's a halo products or not. Lisa Su may have a fiduciary responsibility to charge people as much as the market will bear, but I certainly don't have a responsibility to gladhand them every time they do it. Prices going up due to lack of competition sucks. Prices going up due to dueling duopolies sucks.

Competition is good. When there isn't any, we all lose in some way or another.

And ,yet, i remember during Bulldozer era when Intel jacked up their prices to ridiculous levels , (1000$ for 6c and 1700$ for 8c) ,i guess then was pretty much okay , then.

Yup, new reg, checks out.

NO IT WASN'T.