[ad_1]
The Apple M1 launched earlier this week, and benchmarks for the new chip are already starting to trickle down. While these initial tests are preliminary, they still paint a very positive picture of Apple’s future processor.
In Geekbench 5, an Apple M1 is capable of achieving a score of 1687 single-threaded and 7433 multi-threaded. For comparison, the A14 inside an iPhone scores 1598/3995. This increase in multithreaded performance is likely due to its higher core count (+2 high performance Firestorm processor cores) and the near certainty of higher clock speeds, courtesy of its larger chassis. In the past, iPhone processors have been severely constrained due to high power consumption, so what we see here could represent the A14 finally stretching its legs.
These scores compare extremely well with existing Intel and AMD processors on the market. As always, I recommend caution when extrapolating synthetic tests to actual performance. Synthetics are useful for analyzing low-level functionality and they can shed light on some interesting microarchitectural differences, but we don’t use them as the primary means of evaluating hardware.
The most premium 16-inch MacBook Pro ever sold by Apple features a Core i9-9980HK, an 8C / 16T processor with a base clock of 2.4 GHz and a 5 GHz boost. The measured performance for this processor is 1149/7329 in GB5. Now this is an older 14nm processor. Intel doesn’t currently ship an eight-core Ice Lake mobile processor, but Apple also ships Intel’s ICL-derived quad-core Core i7-1068NG7. This chip obtains a score of 1352/4914. Apple is therefore the leader in single-threaded performance even compared to Intel’s Sunny Cove architecture from 2019, and it is the leader in multithreaded compared to the highest-performing processor cores that Intel delivers to mobile. While Apple only attaches the Core i9-9980HK multi-threaded, the M1’s Geekbench scores clearly beat anything below that point, including the Core i9-9880H.
Since people will be curious about AMD, here’s how it compares. GeekBench 5 clearly likes Zen 3 a lot more than Zen 2; the Ryzen 7 3700X gets a 1281/8220 in GeekBench 5 according to LegitReviews, while the Ryzen 7 5800X gets a 1673 / 11,246. That’s an improvement of 1.3x and 1.36x, respectively, which is well above the gains AMD told us to expect from Zen 2 -> Zen 3 on average (1.19x CPI, and up to 1.25x with IPC and clock). Obviously, however, these processors are both desktop processors. We don’t know how AMD’s Zen 3 mobile cores will compare to its desktop chips, but the Ryzen 4800U gets ~ 1130 in single-threaded GB5 and ~ 6800 in multi-threaded. That’s an average – we’ll be referring to this Lenovo system, in particular, when comparing performance.
To sum up: AMD’s all-new 8-core Zen 3 matches the M1 in single-thread performance and significantly exceeds it in multi-threaded performance, but it achieves both of these results with desktop power consumption. . The data from the comparison suggests that Zen 3 will compare much more favorably than Zen 2, but the data also indicates that AMD’s mobile performance in a 15W envelope is well below that of the allowed desktop power levels, this which only makes sense. If we assume that AMD can improve single-thread performance by 1.3x and 1.36x – meaning they can achieve all of that increase in exactly the same power envelope – our hypothetical Zen 3-based mobile processor obtains a score of 1495/9248. Even if we do these very optimistic assumptions, the AMD mobile system would still consume much more electricity than its Apple counterpart.
Sub-test results
The Intel Core i9-9980HK loses almost all of the single-thread subtests to the Apple M1, but it doesn’t lose them consistently. In some cases, the gap is small; in text compression, the M1 scores 1292, compared to 1177 for the 9980HK. Gaussian blur, face detection, and horizon detection tests all greatly favor the M1, as does the ray tracing test.
The subtle performance compared to AMD’s 4800U is distinct from Intel in terms of which tests Apple wins by which margins, but it isn’t much different in the overall result. AMD’s Ryzen 7 4800U can hit 1318 faces / second compared to Intel’s 750, but the M1 reports 2209. Intel’s 9980HK scores 1218 on the N-body physics test compared to Ryzen’s 936 , but the Apple M1 scores 1769. Interestingly, Intel retains the leadership in the machine learning subtest, scoring 1332 against 1169 on M1 and 965 on 4800U.
Multi-thread or single hole
Multi-threaded testing shows some differences from the ST figures. In ST, the M1 swept Intel and AMD almost everywhere. In multi-threading, it’s much more of a struggle. The M1 wins the country mile AES-XTS compression, but Intel is ahead in text compression, 8284 to 7162 (AMD) to 5528 (Apple). AMD decisively wins the image compression test, 10,392 against Intel at 9,524 and Apple at 7,213. AMD narrowly wins in PDF rendering while the M1 wins in text rendering over the two x86 competitors. The benchmark performance of Intel’s cameras is significantly higher than that of AMD or Apple and decisively wins the rigid body physics test. Intel’s scaling in the machine learning test is terrible, however, and Apple wins this test in multi-threading.
Apple M1 Multi-Threading results, according to GeekBench 5.
What does this tell us about the M1 scaling compared to Intel and AMD chips? It’s not clear. The chip’s performance may still be limited by thermals, or by the split-core configuration Apple uses, or both. The scaling factors for AMD Ryzen 7 4800U, Intel Core i9-9980HK and Apple M1 are 6.0x, 6.38x and 4.4x respectively. Given that Apple works with a thermally constrained environment and a mix of high performance and high efficiency cores, this is not too surprising. This is also why the M1 isn’t just going to wipe out x86’s market position in the next year or two – both Intel and AMD are more competitive in MT than in ST, and loads more and more. work are multithreaded these days. These results are a real threat to x86, but they are not its death rattle, even though they translate well in real world applications.
Performance of the Intel Core i9-9980HK multithreaded subtest, GeekBench 5.
It would be inappropriate to draw any conclusions from the results of a single synthesis test, but nothing in these results implies good things for Intel or AMD. Even though GeekBench 5 performs particularly well on Apple hardware or ARM chips in general, we still see a processor with a nominal 5W TDP taking chips with nominal 15W TDPs and come out with a first winner.
It’s entirely possible that sustained testing under load will reveal that the M1 needs to slow down a bit over time because it lacks a fan, and Apple’s high-end MacBook Pros will always outperform it in real-world testing. This is actually the outcome I would expect, because while Apple takes a lot of heat for its prices, business customers won’t buy $ 2000 to $ 3000 of Intel Core i9 laptops if they can. get better performance in audio and video rendering suites with much less expensive machine.
GeekBench 5 is not a real world benchmark and we will wait for real world testing before drawing any conclusions. But nothing in these results points to weak spots in the architecture, and none of its sub-scores highlight issues that AMD or Intel can immediately exploit. Given that we are looking for a 4 + 4 processor against full octa-core solutions, GeekBench 5 has a lot of room to overestimate Apple’s performance without changing the outcome: the M1 seems, at least, to be very competitive with x86 core- for-core and clock-for-clock. Further testing will determine to what extent this is true, or conversely, establish that the GB5 is too Apple-compatible to be a useful cross-platform benchmark.
With tests like Cinebench R23 now supporting M1, we’ll have this question answered as soon as possible.
Also – just to review the “98% of PC laptop” bit – the multi-threaded subtests actually show why this is is not true. “Faster” has a broad contextual meaning, and the Core i9-9980HK is way above the M1 in Text Compression, Image Compression, SQLite, PDF Rendering, Clang, N-Body Physics, Rigid Body Physics and Sub. -HDR test. Even under a benchmark that shows Apple’s M1 in exceptionally good light, there are some distinct and peculiar workloads where the Core i9-9980HK is not only competitive, but victorious with a sufficiently large margin. so that low-end Intel processors have a good chance. -to-vis-à-vis the M1 also.
Likewise, however, Apple’s over-marketing of this silly claim shouldn’t make anyone ignore this bullet point. This represents the most powerful threat to x86 domination that I have seen in my entire career.
Now read:
[ad_2]
Source link