Monday, September 11th 2023

AMD Ryzen Z1 APU Utilizes Zen 4c Cores - Discovered by Reviewer in China

A die-shot of AMD's 4 nm "Phoenix 2" monolithic APU emerged over the weekend—possibly the first example of a Team Red hybrid core processor, utilizing a combination of bog standard Zen 4 cores as well as "compacted" Zen 4c units. Phoenix 2 has been hiding in plain sight it seems, within Ryzen Z1 series APUs—that have much in common with mobile/laptop-oriented 7040U products. David Huang has posted an analysis of a Ryzen Z1 APU via his review as posted on Zhuanlan, where he investigates the intriguing combination of Zen 4 and Zen 4c cores.

As interpreted/translated by Tom's Hardware: "HWiNFO, a system information, monitoring, and diagnostics utility, confirms that the Ryzen Z1, codenamed Phoenix 2, is on the PHX2-A0 stepping. It differs from AMD's Ryzen 7040U series (Phoenix) with the PHX-A1 stepping. The Ryzen Z1 has been rumored to be a clone of the Ryzen 5 7540U for a long time now." Laptops housing the latter APU are reported to have reached retail markets in certain territories, while the Ryzen Z1 (non-Extreme) SoC has not debuted in any new devices. A cheaper ASUS ROG Ally is expected to arrive in the near future with the lesser chip.
Huang's benchmarks point to the hybrid "Phoenix 2" Ryzen Z1 and Ryzen 5 7540U APUs being better suited for casual gaming, while the traditional "Phoenix" setup with a greater TDP—in the Ryzen Z1 Extreme processor—can handle more modern and graphically demanding titles.
Sources: Tom's Hardware, VideoCardz, David Huang via Zhuanlan
Add your own comment

13 Comments on AMD Ryzen Z1 APU Utilizes Zen 4c Cores - Discovered by Reviewer in China

#1
Denver
This TS score is a little above Vega 7(@1900Mhz) and almost equal to RX 660M (6CU - RDNA2), It seems to me that the ideal number would be 6-8CU for low power Socs.
Posted on Reply
#2
FoulOnWhite
So is this AMD's E core CPU? If so, it was inevitable.
Posted on Reply
#3
Assimilator
"Discovered by Reviewer in China", why does this need to be in the title?
Posted on Reply
#4
Mawkzin
FoulOnWhiteSo is this AMD's E core CPU? If so, it was inevitable.
It's not an E core.
E cores from intel uses a reduced ISA compared to P cores, those Zen4c cores uses the same ISA from normal Zen4 with less L3 cache and dense libraries.
Posted on Reply
#5
FoulOnWhite
MawkzinIt's not an E core.
E cores from intel uses a reduced ISA compared to P cores, those Zen4c cores uses the same ISA from normal Zen4 with less L3 cache and dense libraries.
Same principle though, kind of BIGlittle i guess, with high and low power cores. Makes sense for them to do it.
Posted on Reply
#6
trsttte
I hope this rumour is just that and it turns out it ain't true, in my opinion it would be very shady for AMD to release a heterogeneous cpu without any announcement or public information.

It also doesn't make any sense, why wouldn't they announce it?
Posted on Reply
#7
Denver
trsttteI hope this rumour is just that and it turns out it ain't true, in my opinion it would be very shady for AMD to release a heterogeneous cpu without any announcement or public information.

It also doesn't make any sense, why wouldn't they announce it?
They had already announced the existence of Zen4c, the slides are there.

That's no big deal. With limited TDP, laptop CPUs are already at best with most cores running stable at around 4ghz, only 1 or 2 cores have higher boost. Precisely these two cores are Zen4, as the IPC is identical, there will be virtually no glaring difference in this case.
Posted on Reply
#8
tabascosauz
DenverThey had already announced the existence of Zen4c, the slides are there.

That's no big deal. With limited TDP, laptop CPUs are already at best with most cores running stable at around 4ghz, only 1 or 2 cores have higher boost. Precisely these two cores are Zen4, as the IPC is identical, there will be virtually no glaring difference in this case.
Assuming that Zen 4c as implemented in Phoenix-2 is even capable of hitting 4GHz (on wall power, of course). Nothing is confirmed yet obviously but so far the only reference we have is Bergamo, where Zen 4c is clocked quite a bit lower closer to 3.1GHz. Could be the case that AMD is artificially keeping clocks low to protect its Genoa SKUs, or that's just where Zen 4c is since 1T freq does not present thermal issues.

While on battery of course it doesn't matter, but if that's not the case then there might still be a tangible difference, and it's still kinda shady to be mixing PHX-2 supply in with Phoenix for the same SKU (7540U) as the sources on yesterday's Phoenix-2 article suggested. At some point in the past few months I'm pretty sure I saw AMD's website advertise die size at 137mm^2 for 7540U (which is when the PHX-2 rumors started swirling like mad), but it's since been removed.

I think Phoronix is coming out with a P14s Gen 4 review but their unit is the higher end 7840U version. Really need someone to review the 7540U

I assume there will be no problem with scheduling, however, as long as the two Zen 4 cores are always listed first as Cores 0 and 1 like Intel does, and labelled as #1/#1 or #1/#2 in CPPC as usual.
Posted on Reply
#9
Denver
tabascosauzAssuming that Zen 4c as implemented in Phoenix-2 is even capable of hitting 4GHz (on wall power, of course). Nothing is confirmed yet obviously but so far the only reference we have is Bergamo, where Zen 4c is clocked quite a bit lower closer to 3.1GHz. Could be the case that AMD is artificially keeping clocks low to protect its Genoa SKUs, or that's just where Zen 4c is since 1T freq does not present thermal issues.

While on battery of course it doesn't matter, but if that's not the case then there might still be a tangible difference, and it's still kinda shady to be mixing PHX-2 supply in with Phoenix for the same SKU (7540U) as the sources on yesterday's Phoenix-2 article suggested. At some point in the past few months I'm pretty sure I saw AMD's website advertise die size at 137mm^2 for 7540U (which is when the PHX-2 rumors started swirling like mad), but it's since been removed.

I think Phoronix is coming out with a P14s Gen 4 review but their unit is the higher end 7840U version. Really need someone to review the 7540U

I assume there will be no problem with scheduling, however, as long as the two Zen 4 cores are always listed first as Cores 0 and 1 like Intel does, and labelled as #1/#1 or #1/#2 in CPPC as usual.
In general, all Server CPUs always run at lower clocks... There are 128 cores squeezing in a TDP of 360w, AMD certainly had to lower the clockrate, the efficiency curve in the Zen4c must also be moved down compared to the Zen4, these CPUs are a response not only to Intel but to ARM. So the focus on efficiency is the key point...

But that doesn't necessarily mean it's going to be something like just 3.1Ghz on laptop CPUs. Reviews should give the final answer.
Posted on Reply
#10
tabascosauz
DenverIn general, all Server CPUs always run at lower clocks... There are 128 cores squeezing in a TDP of 360w, AMD certainly had to lower the clockrate, the efficiency curve in the Zen4c must also be moved down compared to the Zen4, these CPUs are a response not only to Intel but to ARM. So the focus on efficiency is the key point...

But that doesn't necessarily mean it's going to be something like just 3.1Ghz on laptop CPUs. Reviews should give the final answer.
I didn't say that wasn't the case for Bergamo like all server CPUs, but the key differentiator for Zen 4c and 4 should be 1T max boost, which has nothing to do with 360W TDP or PPT because only single thread or core is concerned, and not really relevant to thermal or overall power efficiency considerations. The base clock is closer to what matters for max TDP.

In that metric there is a significant gap between Bergamo and Genoa, of 600MHz to 1GHz for 1T.
Posted on Reply
#11
Denver
tabascosauzI didn't say that wasn't the case for Bergamo like all server CPUs, but the key differentiator for Zen 4c and 4 should be 1T max boost, which has nothing to do with 360W TDP or PPT because only single thread or core is concerned, and not really relevant to thermal or overall power efficiency considerations. The base clock is closer to what matters for max TDP.

In that metric there is a significant gap between Bergamo and Genoa, of 600MHz to 1GHz for 1T.
The website that did the review has all this data:

"The Zen4c (dense) cores operate at a different frequency compared to the ‘classic’ cores, with Zen4c running at 3.5 GHz and Zen4 at 4.9 GHz. However, when both types of cores are configured to run at the same 3.2 GHz frequency, their performance becomes nearly identical. This is due to the fact that both core types utilize the same ISA and frontend. An interesting aspect is that four out of six cores are designed to run at 3.5 GHz. The Phoenix2 requires less power to attain its maximum boost clock of 4.9 GHz, with power consumption measured at a mere 12.7W."



Posted on Reply
#12
Zubasa
trsttteI hope this rumour is just that and it turns out it ain't true, in my opinion it would be very shady for AMD to release a heterogeneous cpu without any announcement or public information.

It also doesn't make any sense, why wouldn't they announce it?
Unlike Intel's E-cores, all of the core on this have the same ISA. Meaning it doesn't require some exotic scheduling to work properly.
This can function with just CPPC2 on Win10, the whole "favorite core" thing.
Posted on Reply
#13
trsttte
ZubasaUnlike Intel's E-cores, all of the core on this have the same ISA. Meaning it doesn't require some exotic scheduling to work properly.
This can function with just CPPC2 on Win10, the whole "favorite core" thing.
I know that, but at least from what I understood they won't clock as high as regular cores and a heterogeneous cpu i.e. 2 zen4 + 4 zen4c cores won't have the same performance as a regular processor with 6 zen4 cores. Also as cpu's are announced with maximum boost clocks both for single thread and multithread (si lotery aside) it's shady to have a silent change where the same model number will never clock remotely close to advertised.

They even had that weird decoder wheel for the details of each cpu model number, that didn't include any kind of heterogeneous combination.

I know the differences on power and thermal constrained devices like handhelds and laptops will be tiny but that's not the point, it's like when the likes of adata or kingston change ssd components, some of the revisions are pretty close but it's still super shady
Posted on Reply
Add your own comment
Nov 21st, 2024 12:05 EST change timezone

New Forum Posts

Popular Reviews

Controversial News Posts