News Intel Lunar Lake CPU benchmarks reveal good power efficiency and GPU performance — Core Ultra 7 268V results show regressions in multi-core perform...

[Admin]

One of the first third-party reviews of Lunar Lake is now out. It reveals that the Core Ultra 7 268V features good power efficiency and good single-core performance but very mediocre multicore performance.

Intel Lunar Lake CPU benchmarks reveal good power efficiency and GPU performance — Core Ultra 7 268V results show regressions in multi-core perform... : Read more

 

[baboma]

WccfTech has a more detailed take:

Intel Core Ultra 7 268V "Lunar Lake" Shows Impressive Single-Core Performance In Early Review

Vietnamese YouTuber benchmarks the Intel Core Ultra 7 268V "Lunar Lake" chip, revealing the true potential of the 8c/8t CPU in various tests.

wccftech.com

It's Dell XPS 13 (Ultra 7 LNL) vs Asus Vivobook S 14 (Ryzen 370), so it's not a like-to-like compare. That aside, single-core perf is essentially the same for both (LNL 2054 vs Ryzen 2049), while LNL wins on efficiency and Ryzen wins on multicore. No surprise there.

The most noteworthy finding comes from the gaming benchmarks: While LNL has a slight edge over 370 in TimeSpy at default power levels. The 370 laptop can be cranked up to 55W while the LNL model is stuck at 17W, which means 370 would definitively beat LNL in graphics performance when on AC, both in gaming and heavy graphics workload.


Side note: The reviewer's Huế accent is hard to parse. It would've been nice to have a presenter with northern, or even southern accent. But tech bloggers are who they are.

English viewers can use YT's auto-translate for a better understanding, but the YT feature is pretty poor. This would be a good test for Copilot+'s Live Caption feature. It's a tough task given the reviewer's heavy accent and the tech jargons used.

 

[thestryker]

Regression implies it's worse than what came before, but this is a whole new efficiency SKU from Intel. Comparisons to the 165U would probably be the closest version of a like for like and I imagine the 268V keeps up in MT (might even win when TDP normalized) and wins in ST.

 

[purposelycryptic]

Given that these are intended for "stylish", super-thin ultrabooks whose main purpose is being lightweight and still offering long battery life, I don't think too many of them will be seeing heavy MT workloads.

As for performance at higher power levels, that largely defeats the entire purpose of using one of these in the first place - if you intend to regularly use a machine while plugged in, you don't need an anorexic machine designed around sipping power at the lowest rate possible.

I would never buy any of these ultra-slim extended life "AI" laptops, whether AMD or Intel-powered; and most definitely not ARM. Even when traveling, I never use any of my laptops away from an outlet enough to ever even get close to depleting my battery. And even if I somehow did due to unforeseen circumstances, I always have a hefty power bank in my bag in case my phone or NC headphones need a charge.

So, I am obviously not part of the target demographic for these things, but I do wonder how large the actual market for these really is. These processors, both AMD and Intel, seem to have been created entirely as a reaction to Microsoft going crazy pushing Qualcomm's Windows-on-Arm "AI" PCs, which, in turn, seemed to have been a reaction to Apple's ARM-based MacBooks.

But there isn't all that much crossover between the Mac and PC markets, their userbases tend to be different, with different priorities and needs, and Apple is essentially to the computer market what Nintendo is to the Console market - not really a direct competitor.

There has yet to be any real concrete evidence that a sufficiently large demand for the new Windows-on-Arm paradigm even exists, and, while x86 machines following the same design priorities will clearly be more attractive for anyone who actually understands the difference, all of this seems like a rather massive market shift over a largely unproven concept.

PC people obviously see Mac users with endless battery life and feel a desire to have that as well, but, at least from what I can tell, the majority simply want that because it was something they didn't have, rather than it being something they are willing to sacrifice other attributes they get more value out of for.

It could just be something unique to my extended social environment, but I have yet to meet a PC user who would have any interest in using a MacBook, even for free, and only a quite small number that have shown any interest in a MacBook-type PC.

Anyway, I certainly hope this all works out for everyone - I have a good amount of money in both AMD and Intel, and would very much like to see them both do well. And Intel really needs it at the moment. The last thing I want is to see them bought out by Qualcomm or Nvidia. We really need two big boys focused on the PC processor arena, so they can keep one another in check and keep pushing each other on the innovation front.

 

[Gururu]

Pretty impressive and welcome step in the right direction. Unfortunately for AMD, I foresee no increase in market share this generation.

 

[usertests]

Multi-thread performance doesn't matter that much anymore... in many segments (see the great 16-core "stagnation"). What is interesting is that AMD's lead is very consistent across a wide range of power levels, including 15 Watt. Not only is Lunar Lake being beaten, but AMD could beat it while using substantially less power. This may bode well for the Z2 Extreme.

If it has better single-core and graphics than Strix Point, Lunar Lake has earned its place. Although these leads could be tenuous at best, as seen by the single-core score being virtually identical. If Intel cherry picked hard to come up with Lunar Lake's claimed 16% graphics lead over Ryzen AI 370, that claim will fall apart in reviews.

It should be better than AMD's Kraken Point, but maybe nobody will care if Kraken is enormously cheaper.

 

[heffeque]

Just as a quick comment... Intel tends to do fairly well on GPU benchmarks, but fairly worse on real gaming conditions, so... I'd assume that this is still the case and we'll have to see actual performance to know if it isn't.

 

[rtoaht]

Regression implies it's worse than what came before, but this is a whole new efficiency SKU from Intel. Comparisons to the 165U would probably be the closest version of a like for like and I imagine the 268V keeps up in MT (might even win when TDP normalized) and wins in ST.

Yes, precisely. Arrow Lake mobile is the successor of Meteor Lake and Raptor Lake mobile while the Arrow Lake desktop is the successor of Raptor Lake desktop. Lunar Lake is a new genre to compete with Apple and Qualcomm in power efficiency.

 

[watzupken]

Pretty impressive and welcome step in the right direction. Unfortunately for AMD, I foresee no increase in market share this generation.

But I think you missed the point that price is a serious consideration. AMD’s latest chips are manufactured on 4 nm which is just an updated 5nm with a nicer name. Intel went for cutting edge 3nm and it’s a whole new pricing altogether. So yes, if its really better, is it that much better that people will be will to pay a premium for it with on die memory that you can’t upgrade?

 

[thestryker]

But I think you missed the point that price is a serious consideration. AMD’s latest chips are manufactured on 4 nm which is just an updated 5nm with a nicer name. Intel went for cutting edge 3nm and it’s a whole new pricing altogether. So yes, if its really better, is it that much better that people will be will to pay a premium for it with on die memory that you can’t upgrade?

AMD's cheapest Ryzen AI 9 based laptops start at $1200 and that's for the 365 not the 370 Intel has nothing to worry about price wise (Lenovo's 15" LNL starts at $1280). Not to mention every one of the available Ryzen AI 9 laptops use LPDDR5X and thus aren't upgradeable memory wise either.

 

[hotaru251]

so you lose out by neutering your core count and HT? who da thunk?

Efficiency is literally only benefit really (which is expected when you neuter the cores)

 

[TheSecondPower]

The Ryzen AI 9 HX 370 was roughly 57% faster at the 15W power target, 66% faster at 20W, 60% faster at 28W, and 54% faster at 38W. The Z1 extreme was 14% faster than the 268V at 15W, 46% faster at 20W, 35% faster at 28W, and 45% faster at 38W.

These watt figured rarely measure the same thing. The Phawx when reviewing the Asus Ally (or Ally X?) and MSI Claw tested the Claw at a higher power setting because it matched the Ally's lower power setting's real power consumption. That's almost certainly the case here, and I think the meanings change as the power level increases. At 38W, the 370's lead over the 268 is at it's lowest, but it should be exactly the other way around. At higher power the 370's 12 cores and 24 threads can stretch their legs but still stay near their ideal frequency, while the 268's 8 cores should be venturing well outside their ideal frequency.

In short, you could interpret these results like so (these are made-up figures):

reported	268 actual	370 actual
15W	10W	20W
20W	14W	27W
28W	23W	35W
38W	37W	43W

 

[watzupken]

AMD's cheapest Ryzen AI 9 based laptops start at $1200 and that's for the 365 not the 370 Intel has nothing to worry about price wise (Lenovo's 15" LNL starts at $1280). Not to mention every one of the available Ryzen AI 9 laptops use LPDDR5X and thus aren't upgradeable memory wise either.

Agree. But I believe AMD is just being opportunistic here and charging as much as they can since its new and there's no competition. So it is good to have Intel giving them a strong nudge and keeping their pricing in check. What I am saying is that,

1. The cost of each chip is likely higher for Intel which also means the chances of them undercutting or matching AMD is low. AMD can counter by lowering prices.
2. Memory is not upgradeable in most cases, yes. But for those that need more than 32GB of memory, there is absolutely no way to get it. No manufacturers can go around this problem even if they offer the option to customize.

I think the rumored power efficiency improvement sounds great. But I would wait for the actual testing results. Reason is because after using MTL for more than 6 months, I observed the power efficiency is good only when, (1) your laptop is doing nothing, and, (2) when playing/ streaming videos (which benefited from the video encoder/ decoder being a separate chip). Outside of that, the battery life is better than a Raptor Lake H processor, but still falls behind a 7840U.

 

[thestryker]

Agree. But I believe AMD is just being opportunistic here and charging as much as they can since its new and there's no competition. So it is good to have Intel giving them a strong nudge and keeping their pricing in check. What I am saying is that,

1. The cost of each chip is likely higher for Intel which also means the chances of them undercutting or matching AMD is low. AMD can counter by lowering prices.

I wouldn't be so sure about the costs involved because Intel should be getting more chips per wafer by a fair amount. On package DRAM undoubtedly adds a level of cost, but it also simplifies board design which gives OEMs more room for margins to offset a higher CPU cost.

GPD mentioned the 370 costing double their rate for the 8840 so there probably is some room for AMD to drop prices, but the die is also larger (232.5mm^2 vs 178mm^2) so not necessarily a lot of room. This is probably why they've mentioned Krackan being the more budget option. The lower core count hybrid Phoenix chips were quite a bit smaller than the 8 core ones so it stands to reason these will be similarly smaller.

 

[usertests]

These watt figured rarely measure the same thing. The Phawx when reviewing the Asus Ally (or Ally X?) and MSI Claw tested the Claw at a higher power setting because it matched the Ally's lower power setting's real power consumption. That's almost certainly the case here, and I think the meanings change as the power level increases. At 38W, the 370's lead over the 268 is at it's lowest, but it should be exactly the other way around. At higher power the 370's 12 cores and 24 threads can stretch their legs but still stay near their ideal frequency, while the 268's 8 cores should be venturing well outside their ideal frequency.

Big if true!

 

[rm12]

These watt figured rarely measure the same thing. The Phawx when reviewing the Asus Ally (or Ally X?) and MSI Claw tested the Claw at a higher power setting because it matched the Ally's lower power setting's real power consumption. That's almost certainly the case here, and I think the meanings change as the power level increases. At 38W, the 370's lead over the 268 is at it's lowest, but it should be exactly the other way around. At higher power the 370's 12 cores and 24 threads can stretch their legs but still stay near their ideal frequency, while the 268's 8 cores should be venturing well outside their ideal frequency.

In short, you could interpret these results like so (these are made-up figures):

reported	268 actual	370 actual
15W	10W	20W
20W	14W	27W
28W	23W	35W
38W	37W	43W

No, Intel would never publish that they have 15W configured, and than only use 10W. That would be a 30% lower figure that they could put in their marketing materials. I don't understand how you can come to such a conclusion.

 

[systemBuilder_49]

Deduct 10-15% from Intel timespy performance to get a more accurate picture of Intel gaming performance. Intel still trails RDNA3.5 by a substantial amount (5-10%). During demos in early September they used the most thermally limited Asus hx370 laptop to have a straw man to beat...

 

[systemBuilder_49]

I would never buy any of these ultra-slim extended life "AI" laptops, whether AMD or Intel-powered; and most definitely not ARM. Even when traveling, I never use any of my laptops away from an outlet enough to ever even get close to depleting my battery. And even if I somehow did due to unforeseen circumstances, I always have a hefty power bank in my bag in case my phone or NC headphones need a charge.

So, I am obviously not part of the target demographic for these things ...

Intel has thoroughly trained you to accept their design mistakes! Lugging around a desktop processor and a pile of heavy batteries because Intel is incapable of doing efficient CPU design and good foundry management!

I suspect that most people are through with caring around hot thigh burning Intel power pig laptops with heavy batteries! Some cult members may take a little longer than others to deprogram ... I certainly found the 2019 Intel 6-core MacBook pro to be the worst computer I have ever seen in my lifetime .. 95w was not enought power to keep the laptop alive during streaming video meetings with screen sharing! the next time I buy Intel is never ...

- Asus S16 hx365 owner

 

[TheSecondPower]

I certainly found the 2019 Intel 6-core MacBook pro to be the worst computer I have ever seen in my lifetime

Those 2019 Intel Macs had Intel's longest-running 14nm-Skylake-derived CPUs. Mine barely performed better than the 2014 MacBook it replaced. But my 2021 dirt-cheap Tiger Lake Windows laptop easily outdid it in power efficiency, quietness, and single-threaded performance. 2019 was a bad time for Intel. But a lot of Windows laptops did better because they did what Apple didn't until the next year: provide beefier cooling.

 

[TheSecondPower]

No, Intel would never publish that they have 15W configured, and than only use 10W. That would be a 30% lower figure that they could put in their marketing materials. I don't understand how you can come to such a conclusion.

These figures also sometimes come from the OEM, not always AMD or Intel. The Asus Ally and Ally X even have different meanings to their power levels; 15W on the Ally X uses more power than on the Ally. And as I said 15W on the MSI Claw uses less power than 15W on the Ally.

 

[waltc3]

I'd be interested in learning how the HX 370 multicore performance could have been significantly higher at every power level tested, yet supposedly the 268V beat it n performance-per-watt...

 

[Kamen Rider Blade]

It makes you wonder how much better AMD would be if their Mobile APU's were on the more advanced node and wasn't one node behind Intel.

Intel bought out that Fab Waffer start, so they had the advantage for the time being.

But I doubt they can hold onto that Waffer Allocation forever, so AMD will eventually get a crack at it at some point in the future.

 

[thestryker]

Intel bought out that Fab Waffer start, so they had the advantage for the time being.

But I doubt they can hold onto that Waffer Allocation forever, so AMD will eventually get a crack at it at some point in the future.

Apple is the one who does that, not Intel. AMD chose not to use any of the N3 nodes for client products they weren't forced into it.

 

[TheSecondPower]

It makes you wonder how much better AMD would be if their Mobile APU's were on the more advanced node and wasn't one node behind Intel.

AMD chose not to use any of the N3 nodes for client products they weren't forced into it.

AMD's Zen 5c server parts are made on the TSMC N3B node, so some testing would be possible if a reviewer gets ahold of an expensive server chip.

 

[Kamen Rider Blade]

Apple is the one who does that, not Intel. AMD chose not to use any of the N3 nodes for client products they weren't forced into it.

Or the logical conclusion that Server products like Zen 5C will rake in more $$$ on the new more expensive node. So there's no reason to give the new node to cheap ass client customers who won't pay a premium like Enterprise customers.

Ergo, they get last gens Fab Node.