The same and yet different. Intel Core i5-12400 duel (H0 vs. C0)

Ľubomír Samák

2 years ago

Performance tests

We have a shocking revelation! And now, seriously. Different Core i5-12400 processor steppings differ not only in S-Spec code, but also in behavior in practice. Some of the expectations were confirmed, some were not. More importantly, you can choose the more suitable of the two options. But the best part is that everyone can evaluate the “better” variant differently. The differences across these processors must be negligible? No way…

Core i5-12400 or Core i5-12400?

In our analysis of the Core i5-12400 processor, we wrote, among other things, that it comes in two variants that differ in design, and quite dramatically. One is made from the scrap of more expensive processors, and the other physically contains only what the processor needs to run and nothing extra.

While stepping C0 (also known by the S-Spec code SRL4V) has a large, semi-deactivated chip with eight performance cores and the same number of effective cores, stepping H0 (S-Spec SRL5Y) contains only six “P” cores on the silicon wafer and no “E” cores. Which theoretically seems like a better option? There is probably no simple one-or-the-other answer to this. But common sense says that a processor with a fully functional core may be a bit more powerful. This is because there are no stops for deactivated cores in the ring bus.

In the case of a six-core processor like the Core i5-12400, this means four extra stops (two for inactive P cores and two for clusters with E cores). The ring bus works in such a way that it takes one cycle to pass through one stop. So where there is an extra stop after a core is deactivated, there will be two cycles “between” two cores instead of one. So the first hypothesis is that the performance of the Ci5-12400 H0 will be a hair above that of the Ci5-12400 C0.

From the front you can only recognize by the S-Spec code, from the back you can see the clearly different composition of small components (SMD) in the enclosure between the contacts

We undertook this test comparing different Core i5-12400 designs mainly because we measured a surprisingly high idle power draw of over 25 W for the C0 stepping tested in January. This is three times higher than the Core i5-10400 (Comet Lake), which is two generations older. The native 6-core chip could be especially probably because it contains fewer components (after all, who knows how those non-activated or non-functional parts behave…). For now, it looks like stepping C0 can’t theoretically offer anything to make it more interesting. But there is at least one thing.

Because of the 30 % larger casing (209/160 mm²), there should be more efficient cooling, or faster heat transfer from the chip to the IHS and then to a cooler. This is, of course, assuming the same TIM (thermal interface between the chip and the heat spreader) is used. In both cases, however, it should be the same thickness (or rather thinness) of solder.

The solder joint on processors with larger chips, although therefore not directly on the Core i5, has already shown up in a number of delids. You could also see it on the “little” Core i5, in early photos. While it’s possible that these were pre-production engineering samples (ES), it’s likely that there was no change to this in the end. This is confirmed by the very low temperatures of the Core i3-12100F, which is also made from a smaller piece of dough.

But all this is just a kind of thinking about how it could be. Which of these two variants excels in what will only be shown in the detailed tests of the following chapters. The testing methodology is standard, it just won’t go through all the performance tests, but more or less only those in which, in addition to performance, we routinely measure power draw, clock speeds and temperatures. Primarily it is a comparison of operational characteristics. In the third chapter you will also find memory and cache tests. We measure the power draw honestly, properly, with current clamps directly on the wires (EPS) that power the processor.

The CPU cooler is a Noctua NH-U14S, and we traditionally test in a wind tunnel with two 500-rpm Noctua NF-S12A PWM fans at the intake and an equal number of these fans at the exhaust. The airflow is therefore balanced, neither overpressure nor underpressure.

Measurements are carried out on MSI MAG Z690 Tomahawk motherboard with DDR4 memory – Patriot Blackout (4×8 GB, 3600 MHz/CL18). This is the platform on which we test all the cheaper Alder Lake processors from the classes up to Core i5. For a complete listing of test build components, see this link. But now the main thing – analysis and comparison of the characteristics of the processors, which are identical only externally, according to the parameters.

Performance tests

We didn’t get to test everything in turn according to the standard CPU test methodology, but we got the key stuff. We’ve gone through all the tests, alongside which we also monitor and chart power draw, temperatures and core frequencies. So three computational tests, the same number of gaming tests, but also off-load measurements.

In this chapter and in the following chapters, we leave the results without comment. Read through them, make your own picture, and finally, you can stop in the final chapter, where we will make a comment on the measured results.

Computational performance

Gaming performance

Memory tests…

… and cache (L1, L2, L3)

Note: It was not possible to measure the L3 cache results, at least with our component configuration in AIDA64, the corresponding application windows remained empty. Tested with older versions as well as with the latest one (6.60.5900).

CPU power draw curve

Average CPU power draw

We measure the current directly on the 12-volt cables that power the processor

Achieved CPU clock speeds

v

We have a shocking revelation! And now, seriously. Different Core i5-12400 processor steppings differ not only in S-Spec code, but also in behavior in practice. Some of the expectations were confirmed, some were not. More importantly, you can choose the more suitable of the two options. But the best part is that everyone can evaluate the “better” variant differently. The differences across these processors must be negligible? No way…

CPU temperatures

Conclusion

The performance differences between the Core i5-12400 “H0” and “C0” are negligible, but clearly measurable nonetheless. But the results are somewhat against the theory we described in the introduction of the article. More powerful one, and it can be so titled at gains usually up to 1 %, is the stepping C0, the one with the larger core (and S-Spec code SRL4V). Why this behavior occurs we do not know, but we assure you that it is not a measurement error. All outputs are based on the passage of three repeated measurements, they are, of course, consistent and do not look for error in other settings of the power limits. These are unlimited in both cases.

Achieved all-core boost frequencies are about the same, the only significant difference is within the single-threaded boost. But even those 30 MHz to the detriment of H0 stepping are minor, and especially the level of core clock speeds for single-threaded tasks also depends on the quality characteristics of the particular unit. After testing several Core i5-12400 H0 (but also C0), an interval would surely emerge in which the best results with different values would occur. The all-core boost frequencies won’t diverge anymore, as they didn’t in our tests, there’s always a steady 3990 MHz for either processor variant.

More remarkable than the performance results are the power draw results. Stepping H0 (i.e. the one with the smaller, natively 6-core chip) is always more economical. Under “hard” load (3D rendering and x264 encoding) only marginally, by about 3 W (3 %), but at idle the saving (H0) is up to 42 %. Thus, at least the assumption that the variant with the smaller chip must be more economical off-load came out. And with less than 15 W it is. Oddly it’s even 2W below the Core i3-12100F, which might be a bit of an indication that inactive cores don’t increase efficiency and instead make it a bit worse, at least like this with no load.

The biggest surprise is around “gaming” power draw. Stepping H0 here is 8–23 % more economical compared to C0. And while this almost seems like nonsense, know that we’re not wrong about this either (the C0’s performance is also lower in tests on ithardwrae.pl). The test platform, including power limits, is of course identical. The lower idle power draw certainly accounts for some of the overall lower power draw. The formula the lower the load, the greater the difference in power draw, obviously does not apply to gaming. Towards the maximum load (represented by 3D rendering and x264 encoding), the difference in power draw is significantly lower than the ten watts of the idle Windows OS (or roughly five watts in single-threaded tasks), but in games it’s up to 16 W. Specifically in Shadow of the Tomb Raider. In the real-time strategy game Total War Saga: Troy, however, it’s already “only” 8 W to the detriment of the Core i5-12400 C0. This stepping has a slight performance edge, including games, but the drop in power draw is disproportionately higher. The efficiency of the Core i5-12400 H0 (S-Spec SRL5Y) is thus significantly higher for gaming PCs.

In January’s Core i5-12400 (C0) tests, we had little complaints about the efficiency, compared to the Ryzen 5600X it’s weaker. In computing workstations (although the Ci5-12400 isn’t headed for such builds) this statement remains valid regardless of stepping, but in gaming PCs the Core i5-12400 H0 already has the upper hand over AMD’s competing processor. While this may not be the case in all games (for example, in Total War Saga: Troy the Core i5-12400 H0 ties with the Ryzen 5 5600X), on average the more efficient of the Core i5-12400’s pair of processors performs better in this regard.

If you’re building a PC primarily for gaming and efficiency matters, insist on the Core i5-12400 variant with S-Spec code SRL5Y (stepping H0). It will probably be more difficult to negotiate this variant at larger vendors (“they won’t have time to deal with it”), but most smaller ones will hopefully be willing to accommodate such a request. The S-Spec code can be found in several places. In addition to registering in vendors’ systems, it is on the barcode label and also on the processor’s heat spreader, which is visible through a window on the side of the box. But even if you choose the trial-and-error method, there is a fairly high probability that you will hit it. According to our information, there are approximately the same number of processors in circulation from each stepping.

The biggest disadvantage of the Core i5-12400 H0 is worse cooling. This is the tax for a smaller chip from which heat dissipation is slower. In games and applications that make more intensive use of a single core, the difference can be six degrees Celsius even with a more powerful cooler. The supplied Intel Laminar RM1 cooler will thus be noisier on a Core i5 with a smaller chip. The right choice is therefore, as it happens, about priorities.

TL;DR: The biggest difference between the steppings (H0 and C0) of the Core i5-12400 is in power draw. Stepping C0 is significantly more power efficient both off-load and in games (up to 16W). Under hard load it’s only about three watts to the detriment of C0. But this (C0) variant with a larger chip will get less hot despite the higher power draw, and the supplied Intel Laminar RM1 cooler will be quieter for similar performance. Computing and gaming performance is virtually identical (with differences of around one percent), although the H0 variant always pulls the short end of the stick.

English translation and edit by Jozef Dudáš

Continue: Memory and cache tests