Intel Talks Up Xeon Cascade Lake Performance - Up To 48 Cores, 12 DDR4 Channels

[duby229]

Who's "people"? Is that some scientific method measurable fact? I'm a human, does that mean I'm part of "people"? If so, then you lost already, since I'm interested in the max potential of the CPU, and I'm "people" so...

You're the only one who said they need to cater the benchmarks to someone's needs, so no, it doesn't mean I am. I never said they have to cater the benchmarks to my use case.

I said I want them to show the true max potential of this CPU, given the proper software for it. Even software that doesn't exist yet (because it can be written, by me or someone else). So the benchmarks are fine.

Just imagine how retarded such a way of thinking is. By that logic, any new instruction is completely pointless cause it can't benefit "actual existing programs".

Newsflash: new programs can be written or optimized for a CPU. Before you spend effort optimizing it, though, you need to know WHAT to optimize, and the true potential of a CPU. So benchmarks like these are awesome for that.


Example: They release AVX-1024. No program in existence uses them (obviously), but someone is very interested to find out the potential of these CPUs using AVX-1024 before they rewrite their code for it. WTF is so hard to get?

Another example: You know that in synthetic benchmarks the CPU can reach X flops. You optimize your app and it falls short of it, about 80% of the true potential. That means there's still room to optimize it. You'd never know this without the REALLY USEFUL synthetic benchmarks, since there's no other app that comes even close.

Dummy.

I never said synthetics aren't useful, in fact synthetics that -identify- bottlenecks are specifically useful to application developers. In the case of new Instruction set extensions synthetics should be written in two folds, one using the older set of extensions and a second to establish the new set of extensions. It's -only- in this way that bottlenecks between the two can be identified. In this case a good synthetic would be a subset of some actual load.

[pcxmac]

Go easy Weasel.

[Weasel]

I never said synthetics aren't useful, in fact synthetics that -identify- bottlenecks are specifically useful to application developers. In the case of new Instruction set extensions synthetics should be written in two folds, one using the older set of extensions and a second to establish the new set of extensions. It's -only- in this way that bottlenecks between the two can be identified. In this case a good synthetic would be a subset of some actual load.

Man, I've no idea what you're talking about. This isn't about bottlenecks or profiling. This is about benchmarking the potential of the CPU, not of an app.

Let me use another example where you'll hopefully get it. Suppose Joe uses an app that needs a fast population count. The devs of the app profiled it to hell and optimized it as much as they could, but we're in an era where popcnt instruction doesn't exist yet for x86 CPUs.

Intel show new CPU with popcnt benchmarks, blowing anything that used software implementations to count bits before.

If you use Joe's software right now you won't notice much of a difference since the app doesn't use the new instruction. But for the dev of that app he'll now know to make use of the new instruction and gain a massive speed boost.

So the benchmarks show the dev that YES, IT IS WORTHWHILE to convert to the new instructions and show just how fast the CPU can be (i.e. potential).

[duby229]

Man, I've no idea what you're talking about. This isn't about bottlenecks or profiling. This is about benchmarking the potential of the CPU, not of an app.

Let me use another example where you'll hopefully get it. Suppose Joe uses an app that needs a fast population count. The devs of the app profiled it to hell and optimized it as much as they could, but we're in an era where popcnt instruction doesn't exist yet for x86 CPUs.

Intel show new CPU with popcnt benchmarks, blowing anything that used software implementations to count bits before.

If you use Joe's software right now you won't notice much of a difference since the app doesn't use the new instruction. But for the dev of that app he'll now know to make use of the new instruction and gain a massive speed boost.

So the benchmarks show the dev that YES, IT IS WORTHWHILE to convert to the new instructions and show just how fast the CPU can be (i.e. potential).

Dumbass, CPU's -ONLY- run programs.... If the applications that people actually use don't use newer instructions then it doesn't matter. The -ONLY- way for a developer to know whether an instruction set extension can benefit his app is to benchmark it himself. And that's why good synthetics are actual subsets of actual programs.

You keep talking about the "Potential" of your CPU and that actual programs don't matter at all. And even even that it's not about application developers.... What you have been talking about are programs made to give that end user a "feelz gud" moment... And that's in fact all you're getting. This potential you keep speaking of is entirely a figment of your imagination.

[Weasel]

Dumbass, CPU's -ONLY- run programs.... If the applications that people actually use don't use newer instructions then it doesn't matter. The -ONLY- way for a developer to know whether an instruction set extension can benefit his app is to benchmark it himself. And that's why good synthetics are actual subsets of actual programs.

You keep talking about the "Potential" of your CPU and that actual programs don't matter at all. And even even that it's not about application developers.... What you have been talking about are programs made to give that end user a "feelz gud" moment... And that's in fact all you're getting. This potential you keep speaking of is entirely a figment of your imagination.

You are truly a lost cause.

There's a reason GPUs (and in a sense, CPUs, with SIMD) are measured in FLOPS, not based on some piece of shit "typical workload" of some John Average Doe Jr.

Most (all?) of the stats shown on a CPU or GPU are the absolute max you can get. In effect, this tells you the TRUE POTENTIAL of that CPU or GPU. You can of course, run shit software on it, or software not optimized for it, but don't expect anywhere near that.

Let's say you care about memory transfer performance. Your memory is rated (evil synthetic benchmark) as 20 GB/s. You have programs that copy at 10 GB/s. Clearly, your fucking "typical workload" is not optimal. And if you were to code and come to say 19.2 GB/s, you'd know it's almost the limit so the fact that you know the POTENTIAL tells you WHEN TO STOP.

Otherwise, how the FUCK would you know when the performance is "optimal enough"?!? It doesn't matter how many times you profile, you need to know the POTENTIAL of the CPU so that you know how close you are to it.