Wednesday, January 3rd 2024
Intel Meteor Lake P-cores Show IPC Regression Over Raptor Lake?
Intel Core Ultra "Meteor Lake" mobile processor may be the the company's most efficient, but isn't a generation ahead of the 13th Gen Core "Raptor Lake" mobile processors in terms of performance. This isn't just because it has an overall lower CPU core count in its H-segment of SKUs, but also because its performance cores (P-cores) actually post a generational reduction in IPC, as David Huang in his blog testing contemporary mobile processors found out, through a series of single-threaded benchmarks. Huang did a SPECint 2017 performance comparison of Intel's Core Ultra 7 155H, and Core i7-13700H "Raptor Lake," with AMD Ryzen 7 7840HS, 7840H "Phoenix, Zen 4," and Apple M3 Pro and M2 Pro.
In his testing, the 155H, an H-segment processor, was found roughly matching the "Zen 4" based 7840U and 7840HS; while the Core i7-13700H was ahead of the three. Apple's M2 Pro and M3 Pro are a league ahead of all the other chips in terms of IPC. To determine IPC, Huang tested all processors with only one core, and their default clock speeds, and divided SPECint 2017 scores upon average clock speed of the loaded core logged during the course of the benchmark. Its worth noting here that the i7-13000H notebook was using dual-channel (4 sub-channel) DDR5 memory, while the Core Ultra 7 155H notebook was using LPDDR5, however Huang remarks that this shouldn't affect his conclusion that there has been an IPC regression between "Raptor Lake" and "Meteor Lake."
Sources:
David Huang's Blog, Tom's Hardware
In his testing, the 155H, an H-segment processor, was found roughly matching the "Zen 4" based 7840U and 7840HS; while the Core i7-13700H was ahead of the three. Apple's M2 Pro and M3 Pro are a league ahead of all the other chips in terms of IPC. To determine IPC, Huang tested all processors with only one core, and their default clock speeds, and divided SPECint 2017 scores upon average clock speed of the loaded core logged during the course of the benchmark. Its worth noting here that the i7-13000H notebook was using dual-channel (4 sub-channel) DDR5 memory, while the Core Ultra 7 155H notebook was using LPDDR5, however Huang remarks that this shouldn't affect his conclusion that there has been an IPC regression between "Raptor Lake" and "Meteor Lake."
85 Comments on Intel Meteor Lake P-cores Show IPC Regression Over Raptor Lake?
But there's not enough results out there to draw a serious conclusion that IPC has definitely gone down!
You are just remorseful for something that happened fifteen years ago.
wccftech.com/intel-core-ultra-meteor-lake-cpus-already-seeing-big-performance-efficiency-improvements-with-updated-bios/
But turbo clock speed can be all over the place, as this graph of the 155H frequency in a benchmark from Phoronix shows. *3
The 13700H may be able to maintain its rated boost speed more consistently than the 155H, since it's a higher-power processor. Or there could be a big disparity in the cooling the tested laptops provide. Ultimately I don't think this is a good testing methodology for determining IPC. Perhaps a better method would be to monitor the boost frequency while the benchmark runs, determine the average frequency while it was running, and then divide the score by that average. However, the frequency changes so quickly that it would be hard to get an accurate average. And even if that worked perfectly, it wouldn't address disparities in memory. Testing IPC accurately is very difficult especially with laptop processors. Since several variables can't be controlled, multiple different benchmarks need to be run to get a handle on how each variable affects the score, and only then can you begin to take an educated guess about IPC. Any test that fails to find a way to address every variable (other than IPC) is more conjecture than science.
Attribution
*1 Intel's specs for the 13700H:www.intel.com/content/www/us/en/products/sku/232128/intel-core-i713700h-processor-24m-cache-up-to-5-00-ghz/specifications.html
*2 Intel's specs for the 155H:
ark.intel.com/content/www/us/en/ark/products/236847/intel-core-ultra-7-processor-155h-24m-cache-up-to-4-80-ghz.html
*3 "Meteor Lake EPP Tuning For Greater Performance Or Power Efficiency With Intel Ultra Core 7" by Phoronix:
www.phoronix.com/review/intel-meteorlake-epp/2
My day has been stressful so far, but man does this cheer me up, not because I am happy that a company is releasing a bad product, but because it is just plain funny.
Meteor lake and IMHO Arrow Lake will not be about absolute CPU performance but about advanced packaging, advanced fab nodes and all around performance. If you want max CPU performance going forward, only Zen 5 looks promising on that front.
I think its more this is focused on low-wattage and multimedia-type workloads.
I'm hoping they get their act together and bring a meaningful competitor to both the DDR5-based Ryzen and handheld APU lines, but looks like this ain't that gen.
No hate just facts
Intel knows people will buy anything they put out
That's why they refresh over and over.
This means your frequencies will be all over the place.
Most of the actual hands-on reviews of ML laptops show some decent gains. Not enough for me to switch to a laptop again yet, but pretty good. And the AI stuff, after years of waiting, is finally being used effectively in some apps.
I think what people mean to say is, they don't have a use for a 25W laptop. So they compare ML to 45W large / gaming laptops. At least realize that you're comparing apples to oranges here, because the ML line doesn't have an 45W part.
Within its league it's pretty compelling in the x86 space.
Phoronix covers 8th to 12th gen in regards to performance vs power.
www.phoronix.com/review/intel-whiskeylake-meteorlake
I read up on the tiled arch some time ago, and in general its potential is actually not as efficient nor as powerful as monolithic. It is more for what I call 'industrial engineering', specifically it is about increasing yield on increasingly complex dies along with the ability to do some mix-and-match.
But I basically agree these look like Raptor Lake cores adapted to tiled architecture on ML. The 1-2% losses in single thread are likely due to that tiled architecture when you consider the 1365U is 15W. We're talking about rounding-error percentages here though.
Regardless of all that, ML is currently the best general purpose compute sub-45W x86 chip around when you consider that +60-70% multithread advantage vs comparable RPL SKUs. So within its sphere, it's no laggard. Originally there wasn't supposed to be a Raptor Lake, nor a RPL refresh, nor a Rocket Lake for that matter. They were going to go AL -> Meteor Lake.
In that context, it would be double-digit IPC improvement.
www.phoronix.com/review/intel-gen9-mtl-graphics