Not every Core i5-13400F is the same: Raptor (B0) vs. Alder (C0) lake

Conclusion

In the vast majority of Intel Core i5-13400F processor tests (including ours), you’ve studied the results of the variant built on the Golden Cove cores from the Alder Lake generation. But there is also an iteration from Raptor Lake (with Raptor Cove cores). However, this processor is harder to get hold of, its availability is considerably weaker, but it is obtainable. You’ll find out if it’s worth the extra effort from the comparative analysis.

Conclusion

The higher attractiveness of this (B0/SRMBG) or that (C0/SRMBN) Core i5-13400F stepping should be judged by the load intensity of these processors. At high performance, close to the limit of the maximum, stepping B0 achieved lower efficiency in the tests. Computing performance (in Cinebench R23) is indeed a percentage higher, but at the cost of 14% higher power draw. In a single-threaded load, at one-fifth to one-sixth the power draw, the situation reverses, and the short end is pulled by stepping C0. The Intel 7 Ultra variant is 4% faster while drawing up to 29% less power. At the same time, the clock speed of active cores contributing to the single-threaded performance is the same. With such a low power load, the benefit of a more advanced manufacturing process is clearly most evident in the form of higher B0 stepping efficiency.

At medium, typically gaming loads, stepping C0 is again a bit more efficient. However, this is only by a little bit. Stepping B0 can have up to a 5% advantage in gaming performance (and maybe more, we’ll see when we complete the multi-game tests later). Power draw at comparable performance, such as in Total War Saga: Troy, is virtually identical (54.8 vs. 55.6 W). Efficiency in games is very similar for both steppings.

In memory tests, the result of L2 cache latencies turns out to be quite surprising. We expected that its behavior could be the same as that of the Alder Lake die, since Intel originally locked the 2MB L2 cache to 1.25 MB for this model, so that it would not be any different from stepping C0. However, Aida64 measured significantly lower latency with stepping B0 (and the same outputs are also present in other “B0” processor models). This could mean that different behavior is occurring after all, but a difference of 29% (in favor of stepping B0) doesn’t really seem realistic. That’s too significant a reduction. Aida64 can’t seem to measure latency correctly for some reason, and it’s possible that its algorithm would need some fixing to give relevant results on these processors.

Heat is dissipated faster from the C0 silicon, although it’s the slightly smaller die of the pair. This may be due to differences in the properties of the TIMs or, from the perspective of the E cores, their position relative to the P cores. However, processor to processor this can vary – specially for stepping B0. The latter has double the number of clusters (4) for Core i5-13400F, and only 4 cores out of 16 are active (for C0 it is 4 out of 8). Thus, the temperature of E cores on B0 processors may also depend on the mutual distance from P cores, which have higher performance (with higher power draw, higher waste heat). On the tested processor samples there is a larger difference between the temperature of the P and E cores in stepping C0, where there seems to be less interaction. But with the generally low cooling requirements of Core i5-13400F processors, in practice these things don’t matter that much.

But the power draw at very low load (in “idle”) is nicely equal with these processors, which was not the case with their predecessors. The Core i5-13400F stepping H0 was significantly more efficient than C0 in idle. Now the percentage difference is also higher, but at such low levels it practically doesn’t matter. Whether the processor has 2.10 W (Ci5-13400F stepping C0) or 1.85 W (Ci5-13400F stepping C0) is almost irrelevant. The fact that such low power draw is achieved is fascinating on its own. Older processors, as well as current Raptor Lake models with an open multiplier, have significantly higher idle power draw. This is also because, for example, they don’t drop the multiplier to 4 (400 MHz), as is the case with the Core i5-13400F, and the minimum is 8 (800 MHz). And such a clock speed requires a more aggressive power supply for stable operation.

You will have to decide for yourself whether stepping B0 or C0 makes more sense according to the nature of your work. Everyone has some pros and cons. It depends on which characteristics you attach more importance to.

English translation and edit by Jozef Dudáš

Further reading: 


  •  
  •  
  •  
Flattr this!

Comments (10) Add comment

  1. Exactly what I feared. The B0 stepping just as other Raptors is overvolted. A bit more power draw, and higher temperatures. The same rule as to the rest of Raptor CPUs applies – their efficiency and heat generation disappoint if you can’t undervolt them, which thanks to our beloved Intel is now almost impossible.

  2. It would be very cool if at least some version of this processor overclocked all cores to 4.6(max boost) at least on some motherboards.

    1. True, Intel could push the all-core boost of the 65W Core i5s (including 13500/F) a bit… the leap from Ci5-13600K will remain very significant (and not cannibalize) even after the 500 MHz boost you suggested for Ci5-13400(F).

        1. You’re right. My mistake in the comment, for which I apologize. The 13500F really doesn’t exist (and there is only a 13500). Thanks for pointing it out.

  3. The hashtag #7nm is wrong, this CPU is 10nm, the process is called Intel 7 probably to mislead customers.

  4. Since the CPUIDs of Raptor C0 and Alder C0 are different, there is a possibility that Raptor C0 and Alder C0 are not exactly the same chip.
    In that case, the process rules may not be the same either.
    (Core i5-8400 and Core i5-9400, which used the same chip, had matching CPUID 906EA)
    To make a definitive determination as to whether or not they are the same chip, it is necessary to examine them under an electron microscope.

    Raptor B0 (MAX P8 + E16) CPUID B0671
    Raptor C0 (MAX P8 + E8) CPUID B06F2
    Alder C0 (MAX P8 + E8) CPUID 90672
    Alder H0 (MAX P6 + E0) CPUID 90675

    1. That is true. There could be some revision or respin. Personally I expect it’s more likely to be only minor changes though. It could be a change isolated to some metal layer, but perhaps it could even be just a difference in the microcode version that is present on the ship out of the box (I think the base microcode is different, as an effect the 13th generation processors do not work with BCLK OC unlike the 12th gen versions even if they are based on C0 die, for example).

      There could also be changes to the manufacturing process even within the original Intel 7 technology. There always are “knobs” that can be tweaked. I think the Ryzen 3000 “XT” refresh CPU were said to have some manufacturing process tweaks and changes, even though it was still all within the envelope of the N7 process tech originally used for prior SKUs.
      Something like that could have been done by Intel to these C0 die 13th generation processors, too.

Leave a Reply

Your email address will not be published. Required fields are marked *