Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Intel's New Chimera: Alder Lake

agner.org

Post reply on HN
Print view View original

231–240 of 253 posts

Re: Intel's New Chimera: Alder Lake

#231

Post by paulmd »

Nursie wrote: post #130
paulmd wrote: post #117

Earlier quoted context omitted.

I had a college professor who loaned me one for a project and it was absolutely unusable in windows. Took like 10 minutes to boot to the desktop (although maybe he had viruses, wouldn’t have surprised me, and he certainly didn’t have an SSD in it). Lxde is great though and sure id believe that. I ran lxde on a single-core Phenom low power thing and with a SSD it was marginally usable even with that complete shit hard…

Yeah I think the solid state storage card really made it workable, I got an add-on 64Gb card at no small expense, doubled the RAM (to 2GB!), and linux was generally very happy on there.

100%, SSDs make a massive difference in perceived performance. That laptop (it was an AMD V140, it was the Compaq CQ56 shit-top they sent out to replace laptops after bumpgate) was completely unusable with a HDD in windows but between a SSD and lxde it became marginally tolerable. It certainly wasn't fast but it could browse the web and it could play hardware-accelerated youtube if you used H264, etc.

I actually would not be surprised if, at the extreme low-end like the V140 or the Atom netbooks, that a SSD actually produced measurable increases in CPU performance. Not only do you have iowait which generally doesn't count as "idle" time according to linux (and probably windows) cpu utilization metrics, but even if you account for iowait, swapping off for another task while you iowait likely does have a performance hit in terms of caches that are no longer hot and so on, so performance is likely diminished when you come back to it too. And with no SMT, and only a single thread, context switching does become comparatively much more expensive in that fashion.

Re: Intel's New Chimera: Alder Lake

#232

Post by sliken »

roomey wrote: post #33

Just trying to figure this out. So the low power high power cores do seem to be inspired by the M1. If they can implement that easily in x86 is another thing but if anyone can Intel can I guess. The half prescion fps are obviously aimed at ML but how on earth are the hoping to compete with gpus on this front? Where is the use case on this (genuine question). The complex numbers bit makes me think of quantum circuit s…

Alder lake so far hasn't been anywhere near the efficiency of the M1.

> The half prescion fps are obviously aimed at ML but how on earth are the hoping to compete with gpus on this front? Where is the use case on this (genuine question).

Sure, it's not GPU competition, but still helpful, after all a fair chunk of the computers on the planet to do not have discrete GPUs.

Re: Intel's New Chimera: Alder Lake

#233

Post by sliken »

jeffbee wrote: post #167
layer8 wrote: post #140

Earlier quoted context omitted.

Compared to the M1, Intel is still restricted by the existing memory and I/O interfaces and by x86 compatibility, which means they can’t do certain things the M1 did, like the integrated unified memory (for massive memory bandwidth) and the relaxed memory model. And as the siblings explain, the P/E design is more about power efficiency than about performance, although the increased power efficiency does allow for mor…

The M1 has the exact same memory architecture as an 11th generation Intel Core series mobile CPU.

Not really, the M1 has a relaxed memory model, the 11th gen intel core series has a strict memory model. Generally the relaxed memory model allows better bandwidth and performance out of the same memory system.

Not to mention there's the M1 pro, max, and ultra if you need more memory bandwidth.

Re: Intel's New Chimera: Alder Lake

#234

Post by TaylorPhebillo »

fancyfredbot wrote: post #118

It's strange Agner seems so against this hybrid design despite the fact it's been proven to work so well for ARM. I'd agree this architecture is very difficult to optimize a single application for, but on the other hand he doesn't seem to consider whether it can lead to better performance and battery life in everyday use cases with mixed workloads using less than 100% of available performance.

ARM big little uses the same instruction set on each core, right? You wouldn't have one 8.2 core and one 8.4 core?

Re: Intel's New Chimera: Alder Lake

#235

Post by fancyfredbot »

zaptrem wrote: post #228
fancyfredbot wrote: post #118

It's strange Agner seems so against this hybrid design despite the fact it's been proven to work so well for ARM. I'd agree this architecture is very difficult to optimize a single application for, but on the other hand he doesn't seem to consider whether it can lead to better performance and battery life in everyday use cases with mixed workloads using less than 100% of available performance.

He seems more annoyed about bad software thread management in Windows 11 (and none at all in Linux and MacOS) and that P & E cores have different instruction set support, which messes with apps.

It's weird to be annoyed by the difference in instruction set because by default they do have the same instruction set, and neither support avx512. You have to go to great lengths to enable avx512 on the P cores and as he says only certain motherboards can do it. I'll admit it is a shame that neither core supports avx512 by default though. If he'd said he was upset the gracemont E cores didn't emulate avx512 in microcode then I'd agree completely.

Re: Intel's New Chimera: Alder Lake

#236

Post by toast0 »

spijdar wrote: post #81
paulmd wrote: post #68

Earlier quoted context omitted.

the concept of “e-cores” is a rebranding of Atom. The Atom name is extremely tarnished from the days when it was super slow in-order netbook cores but Intel started seriously revamping the design in 2012-2014 with silvermont and airmont. They have been good for several years now - Goldmont (eg J5005) is between core2quad and Nehalem i5 (non-SMT) level performance despite running lower clocks (ie IPC is higher), and t…

As an aside, I don't think even the early Atom processors were as terrible as their reputation branded them. I had an early Atom-based netbook, and it was totally acceptable... if running Linux (or XP). Kind of like the "Vista debacle", I think Atoms were just unsuited to running Windows 7. Maybe I'm too forgiving or forgetful, but I used that netbook for many years, and it wasn't that slow with Linux. (7 was pretty…

I agree, the Atom processors have tended to be fine. Lots of questionable accompanying technology choices or marketing practices though.

1st gen Atoms with the chipset (945?) that was bigger and hotter than the processors was kind of amusing.

That Intel can't help itself from making confusing product names is pretty inexcusable. If a 'Pentium' is an Atom or not makes a big difference in what the specifications mean, and it's not clear to regular people that a dual-core mainstream Intel is about performance equal to a quad-core Atom (assuming similar age chips).

Laptop makers love to combine atom CPUs with amazingly slow spinning hard drives, and they're usually pretty small too. Windows 10 performance on a spinning drive is trash, and it's even worse when it's a low rpm laptop drive. Doesn't help that they're often as little ram as possible and soldered down (although even with 8GB ram, Windows 10 will flog a spinning drive just sitting at the desktop; who knows what it's doing).

It would be interesting to see a 4p+24e or 6p+16e setup for parallelized loads though. Not sure there's currently a need for that on the desktop though.

Re: Intel's New Chimera: Alder Lake

#237

Post by toast0 »

n7pdx wrote: post #168
adam_arthur wrote: post #22

Earlier quoted context omitted.

AMD beats them because they are a process node ahead. Dunno why every comment implies there's some design gap that leads to M1 or AMD being more powerful or efficient. It's 90% the process node. There is no magic design sauce that makes those chips better. The big/little design change did seem to improve performance quite a bit for Intel, even using the same manufacturing node as before

"Dunno why every comment implies there's some design gap that leads to M1 or AMD being more powerful or efficient." That is because there is a huge design gap. Intel architecture and design teams have always assumed on the inevitability of their own manufacturing superiority and process advantage and thus never had to really consider where to place transistors for the most benefit, or think really hard on better ways…

It also doesn't help that their design pipeline and their process pipeline were tightly linked. When Intel's "10 nm" process pipeline stalled, their design pipeline did as well. Alder Lake/Golden Cove is (or seems) to be the first really new design since Skylake, and it uses a lot of watts to compete. We'll see if they get their grove back with the next couple releases, I guess.

Re: Intel's New Chimera: Alder Lake

#238

Post by karsinkk »

quux wrote: post #7
mrath wrote: post #2

I don't have an idea on CPU design. Can this lead to M1 like performance?

The goal of this kind of design is M1 like energy efficiency, not performance (I don't know if the Alder Lake P cores can outperform an M1 P core.) M1 has a similar architecture with performance and efficiency cores, but so far I haven't heard of it causing problems for software developers. All versions of macOS that support M1 also have a scheduler that is able to move threads between E and P cores nicely.

Is there any reading material that you can share about the M1 scheduler?

Re: Intel's New Chimera: Alder Lake

#239

Post by gonzo »

Calling Gracemont (the E cores) “Atom technology” is a bit weak. They support out of order execution and AVX2.

Gracemont is faster, clock-for-clock than Broadwell and nearly as fast as Skylake running ChaChaPoly-1305, as used in WireGuard, largely due to avx2.

Intel’s benchmarking shows 80% performance peak vs peak of 4C of Gracemont vs 2C4T of Skylake.

Re: Intel's New Chimera: Alder Lake

#240

Post by xmodem »

paulmd wrote: post #68
xmodem wrote: post #56

Earlier quoted context omitted.

Indeed, I do sometimes wonder what it'd be like if they made the entire processor out of E cores.

the concept of “e-cores” is a rebranding of Atom. The Atom name is extremely tarnished from the days when it was super slow in-order netbook cores but Intel started seriously revamping the design in 2012-2014 with silvermont and airmont. They have been good for several years now - Goldmont (eg J5005) is between core2quad and Nehalem i5 (non-SMT) level performance despite running lower clocks (ie IPC is higher), and t…

> Denverton is actually a very interesting platform for the wattage - that’s 16 Goldmont cores in a 32W package

That's pretty neat, but I'm thinking about, say, a 60W-95W package that would make sense in a normal desktop.

How many E-cores could you fit in such a package? For highly-parallel workloads, would it stomp all over anything else on the market? (I'd guess yes)

Post reply on HN