Leaked benchmarks reveal major performance leaps for Intel's 15th-gen Arrow Lake chips

[sambitsaha]

In brief: New leaks have revealed that Intel's 15th-gen Arrow Lake chips show promising improvements in CPU performance over their 13th-gen counterparts. There's also an expected uplift in iGPU performance of over 200% in 3DMark when compared to the current Core i9-13900K flagship.

Even though we're nearly a year away from launch, leaks have emerged showing Arrow Lake's performance advancements over current-gen options. Thanks to Igor's Lab, we can start establishing our expectations for the upcoming Intel processors. The results suggest that the 15th-gen Core chips might pose a significant challenge to AMD in terms of performance.

Courtesy of Igor's Lab

In the leaked benchmarks, the top-end Intel Arrow Lake CPU seems to pack a substantial performance uplift of about 20% over the Core i9-13900K in multi-core benchmarks such as Geekbench. Conversely, the 14th-gen scores are rather disappointing to say the least, being only about 1-3 % faster than the current 13th-gen parts. As for single-core performance, we can expect up to 17% improvement for the Arrow Lake chips over the current Raptor Lake CPUs in specific tests.

While the figures for Intel's 14th gen are not impressive, those CPUs still maintain compatibility with the LGA1700 socket used for the 12th- and 13th-gen processors. However, Arrow Lake will use a brand-new LGA1851 socket when it comes out in 2024. Given the leaks, holding on and upgrading straight to 15th gen looks far more lucrative than upgrading to 14th gen.

Courtesy of Igor's Lab

The biggest surprise lies in the graphics department, where the 15th-gen iGPU exhibits an extraordinary improvement of up to 240% over the current 13th-gen offerings in the 3DMark TimeSpy benchmark and a surge of up to 220% in the WildLife Extreme tests.

These figures are incredibly promising, particularly considering that AMD has traditionally led in iGPU performance. Team Red's historical dominance in iGPUs is why most handheld gaming consoles employ AMD APUs, but given these leaks, AMD might be facing a serious threat from Intel in the near future. This is good news for Intel enthusiasts as well as general consumers, as increased competition encourages companies to strive harder to outdo their rivals, ultimately benefiting the end user.

Permalink to story.

https://www.techspot.com/news/99470-leaked-benchmarks-reveal-major-performance-leaps-intel-15th.html

 

[Neatfeatguy]

Intel: Quick! Have PR get something out there about our next, next gen of CPUs because there is a ton of people laughing at us about our 14th gen that's coming out. Make it looks good, like really good and mark it as "projected performance".

 

[Puiu]

Seems like the rumours of clock speed regression with the new architecture seems to be true if they can't break 20% for single core performance compared to the 13th gen.

Intel might actually fall behind AMD's Zen5 in single core perf if they deliver on the 20-25% IPC + small clock speed increase targets.

 

[fps4ever]

Seems like the rumours of clock speed regression with the new architecture seems to be true if they can't break 20% for single core performance compared to the 13th gen.

Intel might actually fall behind AMD's Zen5 in single core perf if they deliver on the 20-25% IPC + small clock speed increase targets.

Although I agree with your assessment I would certainly welcome higher IPC at less clock speed. Chasing those higher speeds for performance come at a cost of higher power and more heat.

 

[ScottSoapbox]

Better iGPU performance would be welcome in my next Surface Pro (as MS has yet to offer AMD options).

 

[Dimitriid]

Although performance gains are nice, I'd be far more pleased with intel if they made far better gains on efficiency instead of trying to blow AMD out, and you can almost mean that 'Blow out' literally given how much heat some of the top chips produce.

The fact they might not so aggressively pursue top clocks this time I think (But I might be wrong) bodes well for a focus on efficiency. iGPU gains are also rather important going forward seeing how Nvidia intends to do just about nothing to revitalize the dedicated GPU market anyway like it or not you'll have to depend on integrated a lot more.

Now if they can only improve their initial igpu driver launch quality by as much as the performance they might have a winner, but I wouldn't hold my breath I don't think they'll be able to beat the Ryzen chips on integrated graphics but just getting a lot closer would make choosing an intel laptop a much better idea than today imo.

 

[wiyosaya]

I would not be surprised if this was yet another attempt by Intel to mislead its potential customers; in other words, Intel Propaganda.

It will be too late for me, anyway, because I'm building new systems for my home computers - NOW. I'd be loathed to choose Intel in any event since literally everything needs upgraded on the Tik.

 

[DSirius]

On paper Intel is the 1st, at least on printer paper. The problem is that fewer and fewer people use papers now.

 

[takaozo]

That PL1=PL2 = 250W is killing me
Intel 15k is going to need a bigger case.

 

[wiyosaya]

That PL1=PL2 = 250W is killing me
Intel 15k is going to need a bigger case.

Hmmm. That looks like the Intel - RTX 4090 combo case.

 

[sorten]

3%-20% compute performance gains over two generations is pretty disappointing. I don't have much use for the iGPU unless my d-GPU fails me. Even in that situation I tend to have a previous gen video card in my closet.

 

[DSirius]

3%-20% compute performance gains over two generations is pretty disappointing. I don't have much use for the iGPU unless my d-GPU fails me. Even in that situation I tend to have a previous gen video card in my closet.

Intel PR team had a mission to brag about 15th gen processors - on PAPER - and having nothing else better, they brag about iGPU performance of over 200% in 3DMark.

 

[PEnnn]

"Leaked benchmarks"

Intel PR machine is leaking again.....

Yawn.

 

[brucek]

That graphs reads "would notice little to no difference in [my] daily use" to me.

 

[passwordistaco]

You might want to look up "major". I don't think it means what you think it means.

 

[hwertz]

Tasty! I have an 11th gen i3-1115G4 (lowest end i3, dual core) and the GPU is already quick enough to run even modern AAA titles on low or medium settings and most games on high. This is in Ubuntu -- as bad as the Linux Intel GPU 3D support was 5 or 6 years ago, they did a FULL rewrite starting about 2-3 years ago (it's a modern-style "Mesa Gallium" driver now) and they now have excellent performance and compatibility. Getting a 2.5x speed boost? That'd make it quite a spicy meatball for gaming and seriously give the Ryzens a run for their money.

 

[Racklefratz]

Chasing those higher speeds for performance come at a cost of higher power and more heat.

Of course. But that's hardly a new phenomenon.

 

[nitebird]

Seems like the rumours of clock speed regression with the new architecture seems to be true if they can't break 20% for single core performance compared to the 13th gen.

Intel might actually fall behind AMD's Zen5 in single core perf if they deliver on the 20-25% IPC + small clock speed increase targets.

20% multi-core performance, no hints yet on IPC single core. It could be more E-cores, better interconnects, scheduling, etc that is not IPC.

 

[Puiu]

20% multi-core performance, no hints yet on IPC single core. It could be more E-cores, better interconnects, scheduling, etc that is not IPC.

"As for single-core performance, we can expect up to 17% improvement for the Arrow Lake chips over the current Raptor Lake CPUs in specific tests."

In the image you can see Geekbench SC score are 1.09-1.13 (+9-13%).

Even if IPC is ~20%, a reduction in clock speeds could make single core perf be closer to 17% as they state in the article.

 

[HardReset]

"As for single-core performance, we can expect up to 17% improvement for the Arrow Lake chips over the current Raptor Lake CPUs in specific tests."

In the image you can see Geekbench SC score are 1.09-1.13 (+9-13%).

Even if IPC is ~20%, a reduction in clock speeds could make single core perf be closer to 17% as they state in the article.

Where this "reduction in clock speeds" rumour comes? Max turbo for "flagship" CPU (6700K was 2015 part):

6700K: 4.2 GHz
7700K: 4.5 GHz
8086K: 5.00 GHz
9900K: 5.00 GHz
10900K: 5.30 GHz
11900K: 5.30 GHz
12900KS: 5.50 GHz
13900KS: 6.00 GHz

Yes, single core turbo clock speed is getting higher, not lower.

We can expect single core turbo to be over 6 GHz on Arrow Lake that makes IPC improvement to be around 10%, not 20%.

 

[Puiu]

Where this "reduction in clock speeds" rumour comes? Max turbo for "flagship" CPU (6700K was 2015 part):

6700K: 4.2 GHz
7700K: 4.5 GHz
8086K: 5.00 GHz
9900K: 5.00 GHz
10900K: 5.30 GHz
11900K: 5.30 GHz
12900KS: 5.50 GHz
13900KS: 6.00 GHz

Yes, single core turbo clock speed is getting higher, not lower.

We can expect single core turbo to be over 6 GHz on Arrow Lake that makes IPC improvement to be around 10%, not 20%.

I think Moore's Law Is Dead talked about it.

 

[HardReset]

I think Moore's Law Is Dead talked about it

I might accept lower clocks for multi core but for since Intel always promotes single core perf, they want single core perforrmance crown at any price.

Checked that MLID video that's basically rumors only. Big IPC gains etc basic stuff that anyone can generate. Clock speeds are supposed to be lower on mobile is most concrete thing they can tell, not hard to guess tbh.

 

[Puiu]

I might accept lower clocks for multi core but for since Intel always promotes single core perf, they want single core perforrmance crown at any price.

Checked that MLID video that's basically rumors only. Big IPC gains etc basic stuff that anyone can generate. Clock speeds are supposed to be lower on mobile is most concrete thing they can tell, not hard to guess tbh.

Possible it could be mobile only or I could be confusing it with meteor lake.

 

[HardReset]

Possible it could be mobile only or I could be confusing it with meteor lake.

Intel is supposedly be using same architecture cores for both mobile and desktop. Since Meteor Lake will probably be laptop only, lower clocks make sense there.