News Ryzen AI 7 Pro 160 bests previous-gen Ryzen 9 — chip hits Geekbench with three Zen 5 and five Zen 5c cores

We might also be looking at AMD's new "Kraken Point" lineup of processors. That would also explain why it is called the Ryzen AI 7, and not Ryzen AI 3xx.

But since we are dealing with an early sample here, there is no sure way to fully confirm whether this is indeed a Kraken Point SKU.

Moreover, Kraken Point is actually rumored to have a 4 Zen 5 and 4 Zen 5C configuration instead.


The chip also comes with Radeon 870M integrated graphics, a new mid-range RDNA 3.5-based iGPU that will compliment the flagship Radeon 890M launched with the Ryzen 9-based AI 300 series variants.

But since this Ryzen Strix Point chip spotted is an early engineering sample/ES before AMD decided to switch to the new "300" series branding, then most probably the igpu could also be the 880M processor.

Because the 880M igpu was also earlier dubbed as "870M".

Also at Computex most of the companies were showcasing prototype Laptops running early samples of this APU. Even the Ryzen AI 9 365 was previously codenamed as "Ryzen AI 9 165" sporting the 870M igpu.


gSTX01f.jpeg



One strange addition is the L3 cache capacity, which comes in at just 8MB. This could be an error on Geekbench's part, but if this information is correct, it represents a significant reduction in L3 cache compared to previous generation parts which had 16MB

Nothing strange. The Ryzen AI 7 parts will sport less L3 cache. There is no error here. Also make a note that we are dealing with an early engineering sample here.


Adding the 'Pro' nomenclature confirms that AMD will make professional/business variants of the AI 300 series, just as it has done with previous CPU generations

PRO variants were already confirmed before. Expect the Ryzen AI 300 PRO SKUs to be released by October, which would be six months after the Ryzen PRO 8000 series launch.

View: https://x.com/AnhPhuH/status/1808699762377642252


So based on this Geekbench leak, we are looking at the Ryzen AI 7 PRO 360 SKU.
.
 
Last edited by a moderator:
That reminds me of outgoing vs incoming call logs ! This author always uses the term "outgoing" for current CPU/GPU products in his write-up, which of course is incorrect/misleading.
 
I was just digging in the Geekbench entry and the code, and it appears the chip isn't throttling, but despite that the chip had a very low boost clock speed of up to 4.3 GHz.

Also, the Lenovo laptop system's 64 GB of DDR5 RAM was just rated at just 1,870 MT/s, which is indeed very slow. This has also effected the overall scores. So the final retail silicon should perform even better.


LbZwrnw.png
 
  • Like
Reactions: usertests
Are these business and enterprise processors? The "Pro" monicker means we could be looking at some exclusive features for these chips.

Never heard of Kracken Point before anyway though. 😎
 
This can't be any coincidence. The Ryzen AI 7 PRO 360 "Strix" APU has also been spotted within BAPCO's CrossMark benchmark.

And yes, like I highlighted before in my previous post, this chip sports the 880M igpu, and not the 870M as this new Geekbench leak shows, and as the article claims.

FWIW, the Ryzen AI 9 HX 370 chip was also spotted as well. Both were using the LPDDR5-7500 memory configuration. Poor scores by the way !

Ryzen AI 9 HX 370 HP laptop.

Ryzen AI 7 PRO 360 Lenovo laptop.


kxgZGJG.png



5Joocbt.png
 
But since we are dealing with an early sample here, there is no sure way to fully confirm whether this is indeed a Kraken Point SKU.
It was obvious to me that a SKU like this would exist, disabling some amount of cores on each cluster since it's possible to have a bad core on both. I wonder if this is the lowest form of the Strix Point die, or if they could do a 2+4 as well (completely intersecting with Kraken).

Never heard of Kracken Point before anyway though. 😎
Kraken Point is to Strix Point, as "Little Phoenix"/Phoenix2 is to Phoenix/Hawk Point (which ended up being used in the low-end 8300G/8500G desktop APUs). Kraken is a smaller die presumed to have 4+4 cores, and 8 CUs.

AMD has done crazy things like disable 8-core APU dies into 2-cores before (e.g. the Ryzen 3 5125C Chromebook chip). Now that they have more money and confidence, they can just make smaller dies to target more segments so they aren't wasting good dies.

Below Kraken Point, we may see a Mendocino successor "Sonoma Valley" with 0+4 cores. Going from Zen 2 to Zen 5C would be great for that segment.
 
It was obvious to me that a SKU like this would exist, disabling some amount of cores on each cluster since it's possible to have a bad core on both. I wonder if this is the lowest form of the Strix Point die, or if they could do a 2+4 as well (completely intersecting with Kraken).

Well not sure about the final config of this Strix Point variant/Kraken Point, but a 2+4 core config wouldn't be out of equation either.

I doubt this one is the lowest die though, since it sports the 880M igpu, a 12CU part. And as per rumors Kraken Point was supposed to feature only 4 WGPs or 8 Compute Units.

So either this is some other Strix Point variant, and not Kraken Point as we assume, but nothing can be confirmed from this leak.
 
Last edited by a moderator:
  • Like
Reactions: usertests
Kraken Point is to Strix Point, as "Little Phoenix"/Phoenix2 is to Phoenix/Hawk Point (which ended up being used in the low-end 8300G/8500G desktop APUs). Kraken is a smaller die presumed to have 4+4 cores, and 8 CUs.
Thanks for the info. That's sounds interesting. Sibling on the way.
 
Below Kraken Point, we may see a Mendocino successor "Sonoma Valley" with 0+4 cores. Going from Zen 2 to Zen 5C would be great for that segment.

There were listings which actually mentioned that the 4nm ESCHER APUs would be replacing Mendocino instead. But as per other rumors and leaks, Hawk Point "refresh" APUs are going to be codenamed as "Escher" instead.

So, "Sonoma valley" should succeed the low-power Mendocino lineup.

The codename "Escher" was also mentioned in one of the Linux LLM patches, and there was a chatter these would be low power APUs targeting 'power constrained' devices. I can't seem to find that patch link now though.

Also, rumors have pointed out that the next Steam DECK console will use the "Sonoma Valley" APU. Most likely sporting Zen 5C cores made on 4nm node by Samsung.

According to one Gamma0burst leak a new processor with the codename ‘Sound Wave’ is also in the works. The only detail mentioned about it is the possible use of a 3nm manufacturing process.

Not sure where will SOUND WAVE fit in the APU lineup though. We also have a SARLAK codename, which was supposed to be the Strix Halo product, but that may not be the case as some claim.

https://gamma0burst.tistory.com/m/1107

Make a note that the process node info listed here might not be accurate or up-to-date.


TQ7QgFP.png
 
  • Like
Reactions: TechyIT223
BTW, a leaked snippet of an AMD slide showed Sarlak/SAR being mentioned after Krackan Point APU lineup, so this means Sarlak is different than the Strix Halo lineup of APUs.

Family ending in 77h.

Fu9jXFTWAAcik4f
 
Last edited by a moderator:
Also, rumors have pointed out that the next Steam DECK console will use the "Sonoma Valley" APU. Most likely sporting Zen 5C cores made on 4nm node by Samsung.
We have a slide naming Sonoma Valley as a drop-in replacement for Mendocino:
https://videocardz.com/newz/leaked-...raphics-specifications-with-two-compute-units
AMD-MENDOCINO-DIAGRAM.jpg


Therefore, it should have weak graphics (even if more than 2 CUs) and memory controller, and would be completely unsuitable for a Steam Deck 2.

I don't think even Kraken Point with 8 CUs of RDNA3.5 is going to entice Valve. They probably want graphics that is over 2x faster than the current Steam Deck APU with 8 CUs of RDNA2, to get similar 1080p performance as Steam Deck 1 gets in 720p (1080p is 2.25x the pixels).

My conclusion is that Valve will either wait a long time for a suitable APU, or get a new custom APU from AMD. The original Steam Deck chip was intended for Microsoft, not Valve.

According to one Gamma0burst leak a new processor with the codename ‘Sound Wave’ is also in the works. The only detail mentioned about it is the possible use of a 3nm manufacturing process.
Moore's Law is Dead leaked that Sound Wave is an ARM APU being made by AMD. Why? Because investors or some company wants it.

 
Last edited:
Therefore, it should have weak graphics (even if more than 2 CUs) and memory controller, and would be completely unsuitable for a Steam Deck 2.

Lol, that 2 CU is too weak for anything related to gaming graphics. They should better wait for a more powerful APU for deck.
 
  • Like
Reactions: usertests
Here is some off topic AMD processor stuff. My web browser/broadband behaving erratically today, doesn't refresh or show new stuff. Not sure why.

Anyway, Wccftech shared some performance details which they got from some of their internal sources. The upcoming AMD Ryzen 9 9900X 12-Core “Zen 5” CPU shows an up to 20% uplift over the previous gen 7900X with PBO !



The Zen 5 Ryzen chip was first tested by using the default 120W TDP mode, and scored around 33,000 points in the Cinebench R23 multi-threaded test. The Ryzen 9 7900X scored around 29,000 points on average at the default 170W TDP, but had PBO enabled by default.

So, that's a 14% uplift coming from a chip sporting 50W less TDP. Seems efficient, eh ?

But once PBO was enabled the Ryzen 9 9900X scored 34,500 points, which is roughly a 20% performance gain.

But these so-called snythetic benchmarks DON'T tell the whole story, and they are not the measuring rod either. The actual "gaming" performance is going to be vastly different.


X2WErub.png
 
But these so-called snythetic benchmarks DON'T tell the whole story, and they are not the measuring rod either. The actual "gaming" performance is going to be vastly different.
It's tough to care about vanilla Zen 5 gaming performance when it will roughly tie Zen 4 X3D and get slapped around by Zen 5 X3D.


Sounds like the only new Zen 5 X3D "feature" is overclocking support, which is not that interesting compared to what else they could have done. But it could look good if they manage to clock both CCDs the same on the 9900X3D and 9950X3D.
 
Sounds like the only new Zen 5 X3D "feature" is overclocking support, which is not that interesting compared to what else they could have done. But it could look good if they manage to clock

That makes little sense since these X3D chips are a sensitive piece of silicon.

Overclocking X3D parts would result in even more heat since these chips don't have much thermal headroom to begin with. So the clocks will still need to be under some limitations.

Unless AMD has made some changes to the silicon and the architecture itself

Btw this is just a rumour for now.
 
That makes little sense since these X3D chips are a sensitive piece of silicon.

Overclocking X3D parts would result in even more heat since these chips don't have much thermal headroom to begin with. So the clocks will still need to be under some limitations.

Unless AMD has made some changes to the silicon and the architecture itself

Btw this is just a rumour for now.
They're not really that sensitive. It's not like these things are clocked at 2 GHz. Also voltage is/was the problem, not heat.
 
  • Like
Reactions: TechyIT223
Who would have thought that the mobile STRIX HALO's BGA 2077 package dubbed FP11, measures 37.5 mm x 45 mm, matching Intel's' LGA 1700 CPU dimensions.

Even larger than the 25 mm x 40 mm size of the FP8 BGA package the "Strix Point," "Hawk Point," and "Phoenix" mobile processors are built on.

The FP11 package dimensions were confirmed via a Shipping manifest data entry spotted recently, as shown here.


GR4fyTFakAA9yYe



As per AMD's own slide, STX Halo uses FP11.

AMD-Strix-Point-Zen-5-Ryzen-APU-Leak-Rumor-Official-Docs.jpeg
 
Last edited by a moderator:
That's pretty much big socket size for a mobile chip. I guess they need to include a lot of features in the die including the igpu as well.

It wouldn't surprise me if AMD is using the same 4 nm node as it did for "Phoenix," for the I/O die. But a more advanced node is warranted, because of the oversized iGPU, which is expected to feature 20 workgroup processors (WGPs), or 40 compute units (CU), 2,560 sps.