HP Dev One With Ryzen 7 PRO 5850U Competes Well Against Intel's Core i7 1280P "Alder Lake P" On Linux

[coder]

The architecture is the same, but the laptop chips are specially binned - or perhaps they use different tsmc tech/low-power libraries, or just lower the clock frequencies, etc.

They can't really be that binned, since probably most of the APU dies go into laptops. Sure, you can buy APUs (i.e. the G-series CPUs) for desktops, but since most desktop users will be using dGPUs, that can't be the majority of their APU sales.

Regardless, their laptop chips run at significantly less power than their desktop chips do. That's all I meant.

Well, okay but you can get Intel laptops which run in low power envelopes, also. Nothing new, there. They have active power management that reduces clock speeds so the chip stays within the pre-determined power envelope. On that point, we should keep in mind that Linux still doesn't have support for Alder Lake's Thread Director.

The 6900HS will sit at 45W for 8 cores, for example. That's a lot less than any of the desktop Zen 3 chips.

They also defined power envelopes for 35 W and 65 W, also. But, testing seems to indicate there's not much gained by going over 35 W. Perhaps that has to do with AMD's reasons for cancelling Zen 3+ for desktop & server - maybe TSMC N6 doesn't scale up to higher power very well.

Will 8 Intel e-cores run at less than 45W? Probably so, but they're also probably quite a bit slower.

Comparing 8 E-cores vs. 8 Zen 3/3+ cores doesn't make a lot of sense, because that's not how Alder Lake would operate. If you wanted an artificial test to measure their respective perf/W, then what I guess would make sense is to measure performance of each, at several different power limits, and plot the respective data points.

Here's some cool data for the 6900HS, but it's too bad they didn't/couldn't do the same for the i9-12900HK:

You can configure a 6800U to run 8 cores at 19W power, and I suspect that would still beat out 8 intel e-cores in performance. I'm not sure exactly how much power the e-cores take, but its probably not a ton less than that.

Huh? No, the E-cores will run within whatever power limit you define. The question is how efficiently they run within a defined power limit. For that, you get a curve, like the ones above. It's illustrated by this unit-less graph from Intel's Architecture Day 2021, which compares it against Skylake cores:

[grok]

I still don't understand where are all the laptops based on Ryzen 6600U/6800U - HP.com has some but they all start at $2200 which is insane.

I still don't understand why there's not been a single formal professional review of Intel's Alder Lake U CPUs, e.g. 1235U or 1265U.

I'd love to replace my aging laptop based on Core i5 6200U which is simply too slow and totally incapable of running any games but there's nothing for me: Intel XE in Alder Lakes is the same as in Tiger Lakes and it trails even the Vega iGPU found in the mobile Ryzen 5000 series. The Ryzen 6000 RDNA2 runs circles around Intel XE but again, there are zero affordable laptops based on 6600U/6800U and 6000H(S) laptops all come with a discrete GPU which I refuse to have.

Intel CPU + Intel GPU might be decent if the software/OS support and experience is good. Might be able to use the other GPU in a VM if you wanted to.

A laptop with the CPU from the Steam Deck would work if you don't mind it being like half a 6000. This is the chip they should be spamming in cheap hardware maybe.

The real test is going to be when Begamo (128-core Zen 4c EPYC) or its successor goes up against Sierra Forest (Intel's first E-core Xeon).

In the near term, it should be interesting to see what AMD has to offer against Alder Lake N (which is what they're calling the next Chromebook SoC; successor to Jasper Lake). Say what you want about Golden Cove (Alder Lake's P-cores), but Gracemont (its E-cores) is a different animal.

That's the 6nm quad core Zen 2, Mendocino with a GPU further cut down from the Steam Deck. That may be good enough.

[grok]

They also defined power envelopes for 35 W and 65 W, also. But, testing seems to indicate there's not much gained by going over 35 W. Perhaps that has to do with AMD's reasons for cancelling Zen 3+ for desktop & server - maybe TSMC N6 doesn't scale up to higher power very well.

I would say it's about production volume. They need all the 7nm and the 6nm they can get. It would also be a pain in the ass to launch an identical slightly better server CPU, with a short useful life.
If they decide to do 6nm PS5 and Xbox the node will quite busy.

[coder]

I would say it's about production volume. They need all the 7nm and the 6nm they can get.

What else are they using N6 for? Or do you think they just need 100% of it for laptops?

It would also be a pain in the ass to launch an identical slightly better server CPU, with a short useful life.

Yeah, my guess is that Zen3+ desktop & server CPUs would've been introduced late last year, if at all. Not very long-lived, plus they were already doing the 3D cache thing.

If they decide to do 6nm PS5 and Xbox the node will quite busy.

True, but why would they? Cost savings from moving to N6 shouldn't be huge, but maybe combined with power savings -> cheaper cooling solution would be worth it? I'd have guessed they won't refresh until N5 becomes cheaper.

[grok]

What else are they using N6 for? Or do you think they just need 100% of it for laptops?

It will be used a lot on desktop and server starting this year, the N6 is in tiles/chiplets for Zen 4 and RDNA3 (along with N5), also the mid range RDNA3 chip.
Maybe you're right on consoles

[coder]

It will be used a lot on desktop and server starting this year, the N6 is in tiles/chiplets for Zen 4 and RDNA3 (along with N5), also the mid range RDNA3 chip.

Zen 4 CCDs are going to be N5. Not sure about I/O dies... normally, I'd expect them to use cheaper N7, except the GPU will now share that same die.

I'm remembering that CDNA 200-series and RDNA 6400/6500 are N6.

Maybe you're right on consoles

Whatever happens with consoles, they need to find a way to get the volume up. More than 1.5 years after launch, and PS5's are still out of stock, even with inflated pricing and expensive bundles.

[smitty3268]

...

We seem to be talking past one another, because all that stuff is common sense and not really what I was getting at.

Anyway, I'll leave it there other than to point out that when i said the 6900HS ran at 45W I was being exact. It's defined with a 35W TDP, but actually runs at 45W. Similary the U chips run at 19W, even though they have a 15W TDP. Just like the Intel chips have a certain defined TDP, but a different actual power usage.

Oh, and

Huh? No, the E-cores will run within whatever power limit you define. The question is how efficiently they run within a defined power limit.

I was referring specifically to the laptop in this article, and how it's configured to run. Not some hypothetical e-core setup.

[coder]

when i said the 6900HS ran at 45W I was being exact. It's defined with a 35W TDP, but actually runs at 45W.

What's funny is that the 6900HS line is rated for 35 W, but the 6900HX models are rated for 45+ W. I hadn't even noticed that, until now.

However, even such an absolute statement is incorrect. Laptops pretty much all have tuned power limits based on their thermal solution and battery. Here's how Anandtech characterized it, in their 6900HS review:

"As with most laptop processor launches, despite the rated TDP on the official processor listing, it’s up to the OEMs to configure and tune the exact performance to the cooling on each unit. This makes comparisons, aside from simply ‘chip vs chip’, quite difficult, as simply adjusting the processor frequency (rather than any other frequencies) has a direct impact on any IPC or performance-per-watt comparison."

"On AMD’s official specifications for the Ryzen 9 6900HS, it lists the TDP as 35 W: the same specifications as the 6900HX, but at an optimized TDP. The HS means that it can only be used in AMD-approved and codesigned systems that can get the best out of the unit: i.e. it is an ultraportable premium device. That being said, laptop vendors can customize the actual final power limit as high as 80W, with the idea that because they are using an optimized voltage/frequency binned processor, the laptop design that can dissipate that much can extract more sustained performance from the processor, this usually translates into a higher all-core frequency."

Source: https://www.anandtech.com/show/17276...3-plus-scaling
(pages 1 & 3)

Similary the U chips run at 19W, even though they have a 15W TDP.

Again, that's going to depend on the specific implementation as well as whether it's plugged in, the temperature, how long it's been boosting, whether you set it in "Turbo" mode, etc.

Oh, and I was referring specifically to the laptop in this article, and how it's configured to run. Not some hypothetical e-core setup.

Well, then your point is irrelevant, because someone using that laptop, in that configuration, wouldn't end up using only E-cores in a performance-intensive scenario.

I thought you were simply musing about the relative perf/W of the different cores. The distinction is that E-cores (like all of them, really) lose a lot of efficiency near the top end of their range. So, if you want to know whether & where it would make more sense to use Intel Gracemont than Zen 3 cores, you'd have to look at the perf/W curves of each and see whether & where they intersect.

However, if you weren't being serious, then we can just drop it.