Live data from Hacker News

AMD's Strix Point: Zen 5 Hits Mobile

chipsandcheese.com

211–220 of 241 posts

Re: AMD's Strix Point: Zen 5 Hits Mobile

#211
post #206

Earlier quoted context omitted.

> R23 doesn’t have the same SIMD optimizations available for ARM as it does for x86. The single-thread benchmark is SIMD-heavy? Now it just sounds like Cinebench ST is a useless benchmark because it's putting a parallelizable SIMD workload on a single core. In real life you'd always be running those multi-threaded, whereas the reason people care about ST performance is for the serialized branch-heavy spaghetti code t…

Yes, there’s no difference between the single and multi core benchmark other than how many threads get spun up. I’m not sure why you’re trying to equate simd with parallelization. Tbh, a lot of your response seems odd to me because it’s making several incorrect assumptions. You can’t really escape parallelization with how any modern core works, even on a single core. You may have certain operations process concurrent…

SIMD workloads generally imply that you're doing the same operation repeatedly. It's literally in the name; single instruction, multiple data. There are occasional cases where that happens but doesn't parallelize well. TLS is probably a good example because you might have to encrypt a network packet and it's big enough to benefit from SIMD but not big enough that the overhead of splitting it across cores is worth it.

But most of the time if you're doing the same operation repeatedly you'll benefit from using more cores. Even for TLS, the client might not split the individual connection across multiple cores, but the server is going to handle multiple clients at once in parallel. Heavy workloads like video encoding make this even more apparent. In general the things that benefit from SIMD are parallel tasks that benefit from multiple cores.

Compare this with, say, a browser running JavaScript in a single tab. There is nothing to put on another core, you don't know what instructions will be executed next until you get there. This is where people actually care about single-thread performance, and where processors achieve it by using branch prediction etc. But these exercise very different parts of the CPU than SIMD-heavy workloads. The latter can easily fill the execution units of a wide processor that would be stymied by the former.

Re: AMD's Strix Point: Zen 5 Hits Mobile

#212
post #206

Earlier quoted context omitted.

Yes, there’s no difference between the single and multi core benchmark other than how many threads get spun up. I’m not sure why you’re trying to equate simd with parallelization. Tbh, a lot of your response seems odd to me because it’s making several incorrect assumptions. You can’t really escape parallelization with how any modern core works, even on a single core. You may have certain operations process concurrent…

SIMD workloads generally imply that you're doing the same operation repeatedly. It's literally in the name; single instruction, multiple data. There are occasional cases where that happens but doesn't parallelize well. TLS is probably a good example because you might have to encrypt a network packet and it's big enough to benefit from SIMD but not big enough that the overhead of splitting it across cores is worth it.…

This feels like a really absurd stretch of trying to discern SIMD away from stuff like standard integer and float operations.

Maybe in the early 90s, but they’re such a part of processor design that you can’t realistically avoid them.

Especially for rendering, which is matrix math heavy, you’d have to design something completely bespoke to avoid it. SIMD is a natural necessity for rendering with any kind of performance.

And because SIMD is such a part of every mainstream processor, it’s very important that benchmarks show how well they perform.

I also don’t understand why you think a JavaScript runtime wouldn’t use SIMD. V8 can make use of SIMD, whether directly targeted or indirectly via the compiler that compiled the runtime itself.

If you want to stress very specific parts of a processor, then use something like SPEC. Cinebench is meant to be a realistic reflection of production rendering.

Re: AMD's Strix Point: Zen 5 Hits Mobile

#213
post #195
post #140

Earlier quoted context omitted.

> Getting 32% higher performance from 17% more cores implies higher performance per core. I don't disagree that it is higher perf/core. It is simply MUCH worse perf/watt because they are forced to clock so high to achieve those results. > The power measurements that site uses are from the plug, which is highly variable to the point of uselessness They measure the HX370 using 119w with the screen off (using an externa…

lmao he’s citing cinebench R15? Which isn’t just ancient but actually emulated on arm, of course. Really digging through the vaults for that one. Geekbench 6 is perfectly fine for that stuff. But that still shows apple tieing in MT and beating the pants off x86 in 1T efficiency. x86 1T boosts being silly is where the real problem comes from. But if they don’t throw 30-35w at a single thread they lose horribly.

> lmao he’s citing cinebench R15?

It's the only one where they measured the power use. I don't get to decide which tests they run. But if their method of measuring power use is going to be meaningless then the associated benchmark result might as well be too, right?

> Geekbench 6 is perfectly fine for that stuff. But that still shows apple tieing in MT and beating the pants off x86 in 1T efficiency.

It shows Apple behind by 8% in ST and 12% in MT with no power measurement for that test at all, but an Apple CPU with a higher TDP. Meanwhile the claim was that AMD hadn't even caught up on the same process, which isn't true.

> x86 1T boosts being silly is where the real problem comes from. But if they don’t throw 30-35w at a single thread they lose horribly.

They don't use 30-35W for a single thread on mobile CPUs. The average for the HX 370 from a set of mostly-threaded benchmarks was 20W when you actually measure the power consumption of the CPU:

https://www.phoronix.com/review/amd-ryzen-ai-9-hx-370/13

On single-threaded tests like PyBench the average was 10W:

https://www.phoronix.com/review/amd-ryzen-ai-9-hx-370/9

34W was the max across all tests, presumably the configured TDP for that system, derived from the tests like compiling LLVM that max out arbitrarily many cores.

Re: AMD's Strix Point: Zen 5 Hits Mobile

#214
post #212

Earlier quoted context omitted.

SIMD workloads generally imply that you're doing the same operation repeatedly. It's literally in the name; single instruction, multiple data. There are occasional cases where that happens but doesn't parallelize well. TLS is probably a good example because you might have to encrypt a network packet and it's big enough to benefit from SIMD but not big enough that the overhead of splitting it across cores is worth it.…

This feels like a really absurd stretch of trying to discern SIMD away from stuff like standard integer and float operations. Maybe in the early 90s, but they’re such a part of processor design that you can’t realistically avoid them. Especially for rendering, which is matrix math heavy, you’d have to design something completely bespoke to avoid it. SIMD is a natural necessity for rendering with any kind of performan…

> Especially for rendering, which is matrix math heavy, you’d have to design something completely bespoke to avoid it. SIMD is a natural necessity for rendering with any kind of performance.

Well sure, but rendering is a classically parallel operation which is regularly implemented as threaded.

> I also don’t understand why you think a JavaScript runtime wouldn’t use SIMD. V8 can make use of SIMD, whether directly targeted or indirectly via the compiler that compiled the runtime itself.

JavaScript runtimes are executing code, so they'll implement the whole gamut and their execution will depend on what kind of code it actually is. But the common JavaScript code, and the kind presumably being tested in JavaScript benchmarks because it's what people care about, isn't implementing a video encoder using SIMD. It's manipulating DOM objects and parsing short pieces of text input, which is branch-heavy code with lots of indirection and very little use of SIMD if any.

> If you want to stress very specific parts of a processor, then use something like SPEC. Cinebench is meant to be a realistic reflection of production rendering.

Which is kind of my point. Production rendering is going to be threaded and max out all the cores, which is Cinebench MT. "CineBench ST" is measuring something that nobody does in real life and doesn't even really correlate with the things people actually do.

It doesn't represent real threaded workloads (which optimally run on many low-clocked cores, not one high-clocked one) nor real serialized workloads (which are full of conditional jumps and cache misses).

Re: AMD's Strix Point: Zen 5 Hits Mobile

#215

Earlier quoted context omitted.

I'm going to need source on that. What does HX 370 score at 10w?

You're asking for a benchmark result for a CPU which just came out and has a configurable TDP that hardly anybody is going to have set to its lowest value, if they even disclose it, much less have done so in a test against the original M1. If you think a source for that even exists you can provide a link. But the result seems pretty obvious. Even the 7nm Ryzen U-series at 15W (e.g. 7730U) was beating the 5nm M1 on mu…

>But the result seems pretty obvious. Even the 7nm Ryzen U-series at 15W (e.g. 7730U) was beating the 5nm M1 on multi-threaded workloads and the HX 370 is well ahead of both on single-thread performance. Single-thread workloads aren't significantly power limited, so to not be the case the Zen5 HX 370 would have to be slower than the Zen3 7730U on threaded workloads at the same TDP, which seems unlikely.

Again, would like a source on that. Please no Cinebench R23.

Re: AMD's Strix Point: Zen 5 Hits Mobile

#216
post #212

Earlier quoted context omitted.

This feels like a really absurd stretch of trying to discern SIMD away from stuff like standard integer and float operations. Maybe in the early 90s, but they’re such a part of processor design that you can’t realistically avoid them. Especially for rendering, which is matrix math heavy, you’d have to design something completely bespoke to avoid it. SIMD is a natural necessity for rendering with any kind of performan…

> Especially for rendering, which is matrix math heavy, you’d have to design something completely bespoke to avoid it. SIMD is a natural necessity for rendering with any kind of performance. Well sure, but rendering is a classically parallel operation which is regularly implemented as threaded. > I also don’t understand why you think a JavaScript runtime wouldn’t use SIMD. V8 can make use of SIMD, whether directly ta…

Rendering is parallel, yes, but it also makes heavy use of SIMD to accelerate the operations per thread. One does not obviate the other.

In the most trivial case, sure, the JavaScript runtimes won’t compile to use SIMD but there’s lot of cases where they will as well as part of their JIT. I think you’re trivializing how they work.

And back to the main point, it doesn’t really matter if you believe Cinebench reflects real world rendering. The fact is that R23 uses SIMD for x86, but not for ARM. R24 rectifies that. Both R23 and R24 use the same rendering code path regardless of running in single or multi threaded mode.

So using R23 as benchmarks for efficiency and performance will naturally benefit x86 significantly. There’s a reason none of the people who push the “AMD is almost the same” use the fairer benchmark to do so. R24 really highlights the actual discrepancy when both are given a fair playing field.

Re: AMD's Strix Point: Zen 5 Hits Mobile

#217

Earlier quoted context omitted.

You're asking for a benchmark result for a CPU which just came out and has a configurable TDP that hardly anybody is going to have set to its lowest value, if they even disclose it, much less have done so in a test against the original M1. If you think a source for that even exists you can provide a link. But the result seems pretty obvious. Even the 7nm Ryzen U-series at 15W (e.g. 7730U) was beating the 5nm M1 on mu…

>But the result seems pretty obvious. Even the 7nm Ryzen U-series at 15W (e.g. 7730U) was beating the 5nm M1 on multi-threaded workloads and the HX 370 is well ahead of both on single-thread performance. Single-thread workloads aren't significantly power limited, so to not be the case the Zen5 HX 370 would have to be slower than the Zen3 7730U on threaded workloads at the same TDP, which seems unlikely. Again, would…

7730U (7nm) vs. M1 (5nm) for MT:

https://nanoreview.net/en/cpu-compare/apple-m1-vs-amd-ryzen-...

Faster in Passmark MT, basically tied in Geekbench MT, faster in average MT score.

HX 370 vs. M1:

https://nanoreview.net/en/cpu-compare/apple-m1-vs-amd-ryzen-...

Faster in everything, ST and MT. ST difference is significant, MT difference is huge. Obviously this is expected because in this comparison AMD has the process advantage, but the expected thing is indeed what happens.

Re: AMD's Strix Point: Zen 5 Hits Mobile

#218

Earlier quoted context omitted.

"Thin and fanless" aren't that hard, just use any low power CPU. But then people also want fast. Apple does this by buying out TSMC's capacity for the latest process nodes and then taking the performance/efficiency trade off in favor of efficiency, so they get something with similar performance and lower power consumption. But then they charge you $400 for $50 worth of RAM and solder it so you can't upgrade it yourse…

Don't run Windows and you don't need fast. Unfortunately Linux on notebooks is always a dice roll of random features (cam, fingerprint, ..) not working. There is a lot of older hardware running like crap because Windows just bloats up.

So is running MacOS on non-apple laptops or running windows on chromebooks. Preinstalled is another story, of course, you paid someone to make sure all those random features work.

That said, defaults seem often wrong and defaults matter. For instance, I recently got an HP Elitebook with an amd 7840hs because for some reason the _u version was tied to a lower res screen. By default it runs high powered and then the fan is loud enough to be annoying. Set it it to balanced or low power and the fan is inaudible.

Re: AMD's Strix Point: Zen 5 Hits Mobile

#219

Earlier quoted context omitted.

>But the result seems pretty obvious. Even the 7nm Ryzen U-series at 15W (e.g. 7730U) was beating the 5nm M1 on multi-threaded workloads and the HX 370 is well ahead of both on single-thread performance. Single-thread workloads aren't significantly power limited, so to not be the case the Zen5 HX 370 would have to be slower than the Zen3 7730U on threaded workloads at the same TDP, which seems unlikely. Again, would…

7730U (7nm) vs. M1 (5nm) for MT: https://nanoreview.net/en/cpu-compare/apple-m1-vs-amd-ryzen-... Faster in Passmark MT, basically tied in Geekbench MT, faster in average MT score. HX 370 vs. M1: https://nanoreview.net/en/cpu-compare/apple-m1-vs-amd-ryzen-... Faster in everything, ST and MT. ST difference is significant, MT difference is huge. Obviously this is expected because in this comparison AMD has the process a…

Let's not use Passmark MT. Stick to the better benchmarks that are optimized for both ARM and x86. GB5 and GB6, M1 is faster in MT despite having 4 fewer cores. If you can find SPEC scores, that'd be great too.

HX 370 vs M1, what's the perf/watt for SPEC and GB5/6 and Cinebench 2024?

HX370 consumes a lot more power. Hence, there aren't any fanless laptops available for it.

4 years later, AMD's chips still can't work in a fanless laptop.

Re: AMD's Strix Point: Zen 5 Hits Mobile

#220
post #216

Earlier quoted context omitted.

> Especially for rendering, which is matrix math heavy, you’d have to design something completely bespoke to avoid it. SIMD is a natural necessity for rendering with any kind of performance. Well sure, but rendering is a classically parallel operation which is regularly implemented as threaded. > I also don’t understand why you think a JavaScript runtime wouldn’t use SIMD. V8 can make use of SIMD, whether directly ta…

Rendering is parallel, yes, but it also makes heavy use of SIMD to accelerate the operations per thread. One does not obviate the other. In the most trivial case, sure, the JavaScript runtimes won’t compile to use SIMD but there’s lot of cases where they will as well as part of their JIT. I think you’re trivializing how they work. And back to the main point, it doesn’t really matter if you believe Cinebench reflects…

> The fact is that R23 uses SIMD for x86, but not for ARM. R24 rectifies that. Both R23 and R24 use the same rendering code path regardless of running in single or multi threaded mode.

But that's not the issue. Even if Cinebench R23 isn't a valid comparison, Zen5 also faster for Cinebench R24 MT.

Cinebench ST (R24 or R23) turns out to be a silly benchmark, because nobody in real life is going to artificially limit their renderer to one thread, but a renderer limited to one thread is also a bad proxy for real single-threaded workloads.

What it's mostly telling you is how wide the CPU is. Which only matters for real single-threaded code if the CPU can find enough instruction-level parallelism to exploit (which that test doesn't probe), and only matters for multi-threaded code to the extent that the processor can maintain that instruction density without becoming limited by thermals/cache/memory/etc. when the code is running on all the cores, which is the thing the MT benchmark tests.

Post reply on HN