Intel Core Ultra 5 125H Squares off Against AMD Ryzen 7 7840HS in Benchmark Leak

[btarunr]

The Intel Core Ultra 5 125H is designed to be a middle-of-the-market processor SKU from Intel's next generation "Meteor Lake" processor family. It comes with a CPU core configuration of 14-core/18-thread. That's 4P+8E+2L (four performance cores, eight efficiency cores, two low-power island cores), although with a full featured Xe-LPG iGPU that has all 8 Xe cores (128 EU) enabled. The chip is normally rated for a 28 W power envelope, although OEMs such as Lenovo have developed a custom 65 W "power mode," which raises the base power value.

A Chinese PC enthusiast with access to an unreleased Lenovo notebook based on this processor, including Lenovo's 65 W Mode toggle, benchmarked it, and compared it with a notebook powered by an AMD Ryzen 7 7840HS "Phoenix" processor (8-core/16-thread, "Zen 4," Radeon 780M iGPU with all 12 compute units enabled); and another notebook powered by Intel's current middle-of-market chip in the H-segment, the Core i5-13500H "Raptor Lake" (4P+8E, Xe-LP iGPU with 5 Xe cores or 80 EU). The results were a little unexpected. The Xe-LPG iGPU of the 125H is shown beating both the Radeon 780M of the Ryzen, and the Xe-LP iGPU of the i5-13500H, with the highest 3DMark Time Spy and Fire Strike scores in the comparison. The Xe-LPG iGPU is 15% faster than the Radeon 780M in Time Spy, and 6% faster in Fire Strike. It's a whopping 70% faster than the Xe-LP iGPU of the "Raptor Lake" chip in this comparison. Things are shockingly different on the CPU performance front for the "Meteor Lake" chip.



In the Cinebench R20 multi-threaded benchmark, the Ryzen 7 7840HS is 10% faster than the Core Ultra 5 125H. It is 6.5% faster in the Cinebench R20 single-threaded benchmark, which is surprising, given that the "Redwood Cove" P-cores of "Meteor Lake" should come with a higher IPC than the "Zen 4" core of the Ryzen. We're not quite sure what's happening here. One possible explanation is that the enthusiast behind the tests used Lenovo's 65 W mode on all three notebooks, and the Ryzen is somehow able to hold onto its boost frequencies better; or there's a software-level problem preventing the benchmarks from correctly scheduling across all 14 cores on the "Meteor Lake."

View at TechPowerUp Main Site | Source

 

[Daven]

If you actually look at the marketing slides for Meteor Lake, no IPC increase is claimed for the P cores, only the E cores.

Meteor Lake is all about tile packaging, NPU, IGP and E cores. I don’t even think the IPC is going up on Arrow Lake P cores.

The Meteor Lake presentation is very vague on performance claims. Even though TPU claims IPC P core increases in the article text, nothing like that is stated in the presentation slides.

Intel Meteor Lake Technical Deep Dive

Today Intel is taking the wraps off their Meteor Lake Architecture. Our tech preview tells you everything you need to know about Intel's new ideas that will power the company's processors for years to come. Just like AMD, Intel is betting on chiplets, which combine multiple silicon dies into a...

www.techpowerup.com

Maybe Intel engineers said it verbally during the keynote.

 

[fancucker]

1 - 65W is a very peculiar TDP selection, MTL achieves 95pc of its performance at 35W
2 - There is massive variation in laptop cooling solutions, firmware based Freq/Power curves, so again a poor comparison
3 - There is a desperate attempt to spin this as a ST perf regression with no tangible evidence
4 - The IGP is much better and needs an opportunity to stretch its legs

 

[pressing on]

The Meteor Lake CPU is a 4.5GHz part; the 7840HS turbos to 5.1 GHz. The 7840HS is eight core versus the four core plus e-core setup of the 125H - maybe equivalent to six cores. The 185H against 7840HS would be a better comparison, even though the 185H is six cores. Meteor Lake launch is only a few days away, let's wait for the full reviews.

 

[tabascosauz]

If you actually look at the marketing slides for Meteor Lake, no IPC increase is claimed for the P cores, only the E cores.

Meteor Lake is all about tile packaging, NPU, IGP and E cores. I don’t even think the IPC is going up on Arrow Lake P cores.

The Meteor Lake presentation is very vague on performance claims. Even though TPU claims IPC P core increases in the article text, nothing like that is stated in the presentation slides.

Intel Meteor Lake Technical Deep Dive

Today Intel is taking the wraps off their Meteor Lake Architecture. Our tech preview tells you everything you need to know about Intel's new ideas that will power the company's processors for years to come. Just like AMD, Intel is betting on chiplets, which combine multiple silicon dies into a...

www.techpowerup.com

Maybe Intel engineers said it verbally during the keynote.

From later interviews as late as October, Intel staff were saying that architectural changes are possible but remain relatively minor. I don't think it was ever going to be a Skylake>Sunny Cove jump.

Disregarding where the performance stands, Intel badly needs the efficiency increase from the node shrink as shown from the Redwood Cove V-F leak. It's not up for debate, Ryzen is far more usable in the -U and -P segments, when away from the wall, because Intel still can't provide a dynamic enough solution for power and clocks. So if they can make major improvements on that front, I don't see a huge problem here - right now 13th gen -U performance is in the absolute toilet and 13th gen -P is too hot and hungry. Meteor Lake needs to fill that gap where Phoenix -U conveniently sits - going higher for -H Intel can just crank up the Vcore and power like they always do.

Hardware scheduling improvements are always welcome.

Besides, for both AMD and Intel (even more so for 13th gen -H here), laptop specific design and optimization always makes or breaks an otherwise healthy CPU, so these scores mean next to nothing at this time.

 

[bug]

From later interviews as late as October, Intel staff were saying that architectural changes are possible but remain relatively minor. I don't think it was ever going to be a Skylake>Sunny Cove jump.

Disregarding where the performance stands, Intel badly needs the efficiency increase from the node shrink as shown from the Redwood Cove V-F leak. It's not up for debate, Ryzen is far more usable in the -U and -P segments, when away from the wall, because Intel still can't provide a dynamic enough solution for power and clocks. So if they can make major improvements on that front, I don't see a huge problem here - right now 13th gen -U performance is in the absolute toilet and 13th gen -P is too hot and hungry. Meteor Lake needs to fill that gap where Phoenix -U conveniently sits - going higher for -H Intel can just crank up the Vcore and power like they always do.

Hardware scheduling improvements are always welcome.

Besides, for both AMD and Intel (even more so for 13th gen -H here), laptop specific design and optimization always makes or breaks an otherwise healthy CPU, so these scores mean next to nothing at this time.

There are Intel Evo designs around P CPUs. And Intel Evo guarantees (among other things) at least 9.5h battery life. So I don't see how your assertion can be true.

 

[tabascosauz]

There are Intel Evo designs around P CPUs. And Intel Evo guarantees (among other things) at least 9.5h battery life. So I don't see how your assertion can be true.

Have you read, like, any -P review ever? Against Rembrandt and Phoenix-U/HS 12th and 13th gen -P have never been fighting anything other than a losing battle on power, thermals, and batt.

If we're going to use Evo as a crutch, then take the Galaxy Book3 with 13th -P against the 6850U Eliteboom 16 for a head to head on the exact same battery capacity (and an i5 against R7 with 680M, no less).

There is a very real need for Meteor Lake in the thin and light segment. It's pretty plain to see from basically every ThinkPad ever after Rembrandt released - two of the same model is almost universally better served by Ryzen.

 

[bug]

Have you read, like, any -P review ever? Against Rembrandt and Phoenix-U/HS 12th and 13th gen -P have never been fighting anything other than a losing battle on power, thermals, and batt.

If we're going to use Evo as a crutch, then take the Galaxy Book3 with 13th -P against the 6850U Eliteboom 16 for a head to head on the exact same battery capacity (and an i5 against R7 with 680M, no less).

There is a very real need for Meteor Lake in the thin and light segment. It's pretty plain to see from basically every ThinkPad ever after Rembrandt released - two of the same model is almost universally better served by Ryzen.

You know as well as I do laptop reviews are almost as rare as hen's teeth. Point me to some and I'll happily read them.
Till then: https://laptopmedia.com/laptops-with-the-best-battery-life/

 

[AnarchoPrimitiv]

Well, there have been a lot of leaks coming out concerning OEMs being completely disappointed in the performance of Meteorlake, specifically that there is almost no increase in IPC, to the point that OEMs have been told to specifically talk about AI first and foremost and CPU specific performance last and least, and this seems to corroborate that I suppose. Obviously, I'll wait for reviews, but if Meteorlake was going to be some significant jump in performance, I think we would have heard about it by now, or at the very least, whispers and rumors of such a performance increase.

FYI, I think notebookcheck.net is one of the best websites for laptop and mobile SoC/GPU reviews and benchmarks, they have a huge amount of data, the ability to directly compare any piece of hardware to another, and often have had benchmark data on mobile CPUs and GPUs or Laptop models I haven't found anywhere else.

 

[kapone32]

You know as well as I do laptop reviews are almost as rare as hen's teeth. Point me to some and I'll happily read them.
Till then: https://laptopmedia.com/laptops-with-the-best-battery-life/

The problem that Intel has is that AMD's is so far ahead a 15 watt part can sell handhelds like hotcakes. Just watch an ETA Prime channel video on handhelds and mini PCs. This is why most AMD centric enthusiasts are chomping for an RDNA2 APU on the desktop.

 

[Daven]

Well, there have been a lot of leaks coming out concerning OEMs being completely disappointed in the performance of Meteorlake, specifically that there is almost no increase in IPC, to the point that OEMs have been told to specifically talk about AI first and foremost and CPU specific performance last and least, and this seems to corroborate that I suppose. Obviously, I'll wait for reviews, but if Meteorlake was going to be some significant jump in performance, I think we would have heard about it by now, or at the very least, whispers and rumors of such a performance increase.

FYI, I think notebookcheck.net is one of the best websites for laptop and mobile SoC/GPU reviews and benchmarks, they have a huge amount of data, the ability to directly compare any piece of hardware to another, and often have had benchmark data on mobile CPUs and GPUs or Laptop models I haven't found anywhere else.

See my post above about IPC. Somehow sites like TPU got it into their minds that Meteor Lake P cores have a huge IPC increase. I don’t think this was ever the case and I don’t know how these sites started believing this unsubstantiated claim. I’m not even sure if Intel outright or even hinted at a P core IPC increase. Maybe with all the Meteor Lake changes including a ‘new’ P core based on efficiency, these sites just assumed that also included higher IPC.

Edit: TPU claims ‘serious’ P and E core IPC increases in the hardware updates articles without reference.

Upcoming Hardware Launches 2023 (Updated Feb 2024)

This article serves as a continuously updated summary of currently known leaks and official announcements regarding upcoming hardware releases in 2024 and beyond. We cover and keep track of developments for Intel Arrow Lake, AMD Zen 5, NVIDIA's and AMD's upcoming GPUs, DDR6 and GDDR7 memory...

www.techpowerup.com

I’m guessing there was some rumor awhile back or a now defunct Intel claim about IPC increases in Meteor Lake P cores but I still don’t know the original source.

 

[R0H1T]

If you actually look at the marketing slides for Meteor Lake, no IPC increase is claimed for the P cores, only the E cores.

So what's Intel doing these days? Tick tick tock tock ~ little to no IPC improvement from 12-14th gen & barely anything on mobile/laptops either!

 

[phanbuey]

So what's Intel doing these days? Tick tick tock tock ~ little to no IPC improvement from 12-14th gen & barely anything on mobile/laptops either!

They're mostly focusing on foundries, but the redwood cove core design in general is underwhelming.

 

[Daven]

So what's Intel doing these days? Tick tick tock tock ~ little to no IPC improvement from 12-14th gen & barely anything on mobile/laptops either!

What’s worse is that Intel could even roll back IPC gains on the P cores in order to increase efficiency and/or compensate for the addition of NPUs and better GPUs. This is not without precedent given that Intel dropped AVX512 from client chips and reduced core counts going from Comet Lake to Rocket Lake.

To be fair, looking at just the P cores nowadays no longer makes sense even if it makes us enthusiasts sad. The whole package has ‘increased computational’ abilities if you look at everything: NPU, P cores, E cores and iGPU. Even the IO tile has cores now.

 

[usiname]

See my post above about IPC. Somehow sites like TPU got it into their minds that Meteor Lake P cores have a huge IPC increase. I don’t think this was ever the case and I don’t know how these sites started believing this unsubstantiated claim. I’m not even sure if Intel outright or even hinted at a P core IPC increase. Maybe with all the Meteor Lake changes including a ‘new’ P core based on efficiency, these sites just assumed that also included higher IPC.

Edit: TPU claims ‘serious’ P and E core IPC increases in the hardware updates articles without reference.

Upcoming Hardware Launches 2023 (Updated Feb 2024)

This article serves as a continuously updated summary of currently known leaks and official announcements regarding upcoming hardware releases in 2024 and beyond. We cover and keep track of developments for Intel Arrow Lake, AMD Zen 5, NVIDIA's and AMD's upcoming GPUs, DDR6 and GDDR7 memory...

www.techpowerup.com

I’m guessing there was some rumor awhile back or a now defunct Intel claim about IPC increases in Meteor Lake P cores but I still don’t know the original source.

Maybe this is the source:
Meteor Lake - Zen 4 killer, eficiency monster
Arrow Lake - 30-40% faster ST than Raptor Lake, (the last rumours point at 5% IPC increase)
Panther Lake - 30-40% faster ST than Arrow Lake, 90-95% faster ST than raptor lake in 2 years

 

[unruheherd]

1 - 65W is a very peculiar TDP selection, MTL achieves 95pc of its performance at 35W

That is what I have intended to say. MTL may fail to impress at 65 W and more, but first leaks have shown a lot of potential at roughly half the wattage and less. I wanted to see a more powerful CPU as well, but I think MTL is a solid base for chips to come. The way Intel is putting together chips and connecting them looks promising. If their forecast holds true and new cores will release on smaller nodes, the MTL successors could become an interesting package at 65 W and above. Although I would prefer to not exceed 95 W for the sake of efficiency, but that is purely my preference.

 

[ThrashZone]

So what's Intel doing these days? Tick tick tock tock ~ little to no IPC improvement from 12-14th gen & barely anything on mobile/laptops either!

Hi,
Intel locking H chips and way over priced HX chips pretty much made my choice for me
Which is amd 7840hs on an acer 17 nitro with 4060.

 

[pressing on]

To be fair, looking at just the P cores nowadays no longer makes sense even if it makes us enthusiasts sad. The whole package has ‘increased computational’ abilities if you look at everything: NPU, P cores, E cores and iGPU. Even the IO tile has cores now.

I see that wccftech have some claimed bench mark results for the Meteor Lake Ultra 9 185H. Like the 13800H this is a 6P + 8E part, top speed is 5.1 GHz versus the 5.2 of the 13800H. The 185H is said to have been in a Samsung Galaxy Book 4 Ultra. The graph below shows these bench marks which are CPU-Z single-thread and multi-thread scores.

The multi-thread score of the 185H is a big step forward from the 13800H. But even allowing for the slight speed difference between the two, I can't see any sign of a significant increase in single-thread performance if these bench marks are correct.

 

[Daven]

I see that wccftech have some claimed bench mark results for the Meteor Lake Ultra 9 185H. Like the 13800H this is a 6P + 8E part, top speed is 5.1 GHz versus the 5.2 of the 13800H. The 185H is said to have been in a Samsung Galaxy Book 4 Ultra. The graph below shows these bench marks which are CPU-Z single-thread and multi-thread scores.

View attachment 325157

The multi-thread score of the 185H is a big step forward from the 13800H. But even allowing for the slight speed difference between the two, I can't see any sign of a significant increase in single-thread performance if these bench marks are correct.

Nice additional results. This would backup the idea that the E cores have increased IPC but not the P cores.

I think Intel is readjusting transistor real estate amongst the tiles to increase efficiency and add new capabilities at the detriment of increased single threaded CPU performance.

 

[tabascosauz]

You know as well as I do laptop reviews are almost as rare as hen's teeth. Point me to some and I'll happily read them.
Till then: https://laptopmedia.com/laptops-with-the-best-battery-life/

I'm sorry, I can't take seriously a "top battery life list" that thinks all laptop batteries come in an identical Wh capacity. Aforementioned comparisons are not at all "rare" in the past 2 years, the state of -U and -P is hardly a secret considering 13th gen changed nothing in this area.

https://www.notebookcheck.net/Samsung-Galaxy-Book3-Pro-16-review-Office-laptop-with-Core-i5-power-and-a-long-battery-life.715024.0.html

https://www.notebookcheck.net/HP-EliteBook-865-G9-Laptop-review-1000-nits-Sure-View-display-not-quite-up-to-par.675587.0.html

https://www.notebookcheck.net/Lenovo-ThinkPad-T14-G3-review-Business-laptop-is-worse-with-Intel-and-Nvidia.702431.0.html

https://www.notebookcheck.net/Lenovo-ThinkPad-T14-G3-review-Business-laptop-is-better-with-AMD-Ryzen-Pro.657465.0.html

https://www.notebookcheck.net/Lenovo-ThinkPad-X13-G3-laptop-review-Endurance-in-outdoor-use-with-Intel-Alder-Lake-U.670955.0.html

https://www.notebookcheck.net/Lenovo-ThinkPad-X13-G3-AMD-Laptop-Review-Already-very-good-ThinkPad-even-better-with-Ryzen-6000.728756.0.html

https://www.notebookcheck.net/Lenovo-ThinkPad-T14s-G4-Intel-Laptop-Review-OLED-instead-of-battery-life.758333.0.html

https://www.notebookcheck.net/Lenovo-ThinkPad-T14s-G4-review-Business-laptop-is-better-with-AMD-Zen4.763581.0.html

ffs you don't need to look past entry #1 on the laptopmedia list. Between 2 identical X1 Gen11s with 57Wh battery, subbing in the 1365U for the 1370P cuts battery life in half. It's the textbook -U vs -P example that has played out everywhere else: no perf vs no battery, your choice.

Meteor Lake can only be a positive in this TDP segment, so I'm not sure why you're defending a thoroughly terrible product against it?

 

[bug]

Meteor Lake can only be a positive in this TDP segment, so I'm not sure why you're defending a thoroughly terrible product against it?

Yes, yes, performance parts will eat battery faster than low-power parts. You just rocked my world.
You said P parts are power hogs, I simply asked how much of a power hog can they be if they can offer 9.5h+ of battery life in Intel Evo designs?

 

[tabascosauz]

Yes, yes, performance parts will eat battery faster than low-power parts. You just rocked my world.

We're pretending that -P is considered "performance" now? LOL

-P exists from 12th gen onwards solely because from 12th gen onwards -U is completely inadequate against Ryzen -U. Doubling power draw across the board (TDP and actual) to stay "competitive" is considered a given now?

Or are you just going to keep fixating on the 9.5Wh number, which doesn't actually address the whole point about efficiency in any way?

 

[bug]

We're pretending that -P is considered "performance" now? LOL

-P exists from 12th gen onwards solely because from 12th gen onwards -U is completely inadequate against Ryzen -U. Doubling power draw across the board (TDP and actual) to stay "competitive" is considered a given now?

Or are you just going to keep fixating on the 9.5Wh number, which doesn't actually address the whole point about efficiency in any way?

What can I say, when a supposedly inefficient mobile CPU gives me 10h battery life, I'm good.

 

[Tek-Check]

Is there anything impressive about Meteor Lake, apart from general iGPU catch-up?

 

[pressing on]

Is there anything impressive about Meteor Lake, apart from general iGPU catch-up?

For some people that alone would be impressive, bearing in mind that a few years back Intel iGPUs could do little more than put an image on screen.