Discussion Intel current and future Lakes & Rapids thread

[Thunder 57]

Touche, but thankfully I don't look like that guy. Whew.

I was being a smartass, you were certainly correct. I just couldn't help myself since I saw that meme recently. Glad you found it in good taste!

 

[DrMrLordX]

I was being a smartass, you were certainly correct. I just couldn't help myself since I saw that meme recently. Glad you found it in good taste!

You could also have used the comic book guy from the Simpsons. Either one works.

 

[Doug S]

I don't think I communicated my question clearly.
Why would Intel NOT be making a clean break if that is clearly the right path?

Because "the right path" will always be a matter of opinion. Intel has gone so long (as long as the "IBM PC" has existed) holding onto obsolescent compatibility that I think there is some reluctance to go all the way when they were finally willing to consider letting go of the legacy cruft. The plan reads like a compromise between those who want to go all the way like I think they should, and those who are resistant to any change and think they should at most pick and choose only those things that ZERO people have needed for the past decade.

Maybe there are niche cases I'm not considering, but unless those niche cases ALSO require modern x86 64 bit performance they can and should be served by existing CPUs (a few SKUs of which will remain available decades from now due to government contracts etc.) and not continue on zombie fashion in future x86 CPUs forever.

 

[Schmide]

That's a really poor half hearted attempt to get rid of legacy. If you're going to do it, really do it!

I can't wait to see Linus tear them a new one when he gets a load of this mess. All support for 16 & 32 bit code should have been removed, including at ring 3. Deprecating instructions but leaving them "ignored" means the opcodes can't be reused which is a real shame for an ISA starved for shorter instruction encodings! Requiring paging in 64 bit mode and booting in 64 bit mode? How's that supposed to work, is UEFI going to have to include a page table? Why not just support a pageless mode, not everything is a PC and not everything needs paging. Or is Intel giving up on x86 for embedded uses and just conceding that to ARM/RISC-V? I guess so.

Because "the right path" will always be a matter of opinion. Intel has gone so long (as long as the "IBM PC" has existed) holding onto obsolescent compatibility that I think there is some reluctance to go all the way when they were finally willing to consider letting go of the legacy cruft. The plan reads like a compromise between those who want to go all the way like I think they should, and those who are resistant to any change and think they should at most pick and choose only those things that ZERO people have needed for the past decade.

Maybe there are niche cases I'm not considering, but unless those niche cases ALSO require modern x86 64 bit performance they can and should be served by existing CPUs (a few SKUs of which will remain available decades from now due to government contracts etc.) and not continue on zombie fashion in future x86 CPUs forever.

Why would you want to clear everything? As it is now the extra modes do not bloat the opcodes, they just set the default operand size and require more specific sizes to go through the REX/VEX prefix.

x86 is not going to give up its robust decoder as it has to support all the vector extensions and the like.

I do wonder if this has to do with all the accelerators intel is adding to the mix? In the end, they're probably just trading one bloat for another.

The bigger question is how much ARM/RISC is willing to add to become competitive? stay competitive?

 

[A///]

This got me though Uni.
Each time i would ask the 8 ball, will my teacher suck... and unfortunately, it always came out "outlook not so good".
I knew i shouldn't even take that class then.

But even then that still got me to graduate....
Intel can't even do that properly as of late.

I used mine to decide whether to talk to a lass or not in university. sometimes it made the wrong decision. how do I know this? her handprint was emblazoned on my face for a good minute. My suave suburb and old country ways were too much for those city folks!

lol thinking back when everything was moved to UEFI.

That was rough but as i recall it was mainly due to people trying in place upgrades. windows 7 began the trend of being able to change components without a reinstall. right around the time uefi became more common place.

 

[Doug S]

The bigger question is how much ARM/RISC is willing to add to become competitive? stay competitive?

What does x86 have that ARM lacks? I can't imagine what you think they need to "add".

 

[Schmide]

What does x86 have that ARM lacks? I can't imagine what you think they need to "add".

vector permute

SVE and the like are great extensions. I guess they can do most of the things AVX can do. Maybe more. The point is extensions are always being added.

 

[soresu]

vector permute

SVE and the like are great extensions. I guess they can do most of the things AVX can do. Maybe more. The point is extensions are always being added.

SME (Scalable Matrix Extension) and SME2 are also complementary compute instruction sets, plus there is TME (Transactional Memory Extension) that was announced at the same time as SVE2 in 2019, hopefully they did not make whatever mistakes Intel did with theirs.

 

[eek2121]

What does x86 have that ARM lacks? I can't imagine what you think they need to "add".

Socketable processors/RAM. A desktop ecosystem (Windows desktop, driver support for GPUs, etc), Desktop class performance at a competitive price point, etc… 🤣

 

[A///]

yeah i also like to ignore companies like ampere existing.

 

[DrMrLordX]

yeah i also like to ignore companies like ampere existing.

Ampere has been quiet lately. Not that this is the correct thread for discussing them, but it's at least somewhat forgivable that people forget about them.

 

[A///]

Ampere has been quiet lately. Not that this is the correct thread for discussing them, but it's at least somewhat forgivable that people forget about them.

incorrect. they've been regularly in the news. it doesn't get picked up because few are using arm power houses in their workflow outside the dc for the most part. they're not the only socketable system that mirrors an x86 setup for the most part. it's a real niche industry at the moe. their 192 tramples the 9654 while using less power. it's very much if you gotta ask you can't afford it type of deal. expanding the reason you don't see the reporting is no one really cares outside of big strides and product highlights. it's even rarer for anyone to get their hands on it vs other used or new hardware not meant for joe shmo consumers.

 

[nicalandia]

I told you so....

 

[A///]

I told you so....

View attachment 80915

welcome to last night mate.

let's see if i did this right.

https://twitter.com/x/status/1660556099660005377

Another one:

https://twitter.com/x/status/1660578093906685952

Maybe MTL-S 6+8 didn't make sense because it's too close to ARL-S and Intel 4 capacity isn't big enough for mobile+desktop?

 

[A///]

6+_8 for arrow lake is interesting. either intel cannot make an 8+16 part, which is likely the reason or their new 6+8 is powerful enough to blast past a 16 full fat cores processor for a total of 14 cores and 20 threads. this isn't the likely reason giving their past performance and behaviour.

if it ain't the 2nd reason they are so screwed for longer than most of us thought. I gave them benefit of the doubt. we were all so close to getting funk pops made of Pat "Guns Ablazing Gunslinger" Gelsinger made for the vast number of intel believers.seems a grand waste now.

 

[eek2121]

yeah i also like to ignore companies like ampere existing.

Where can I buy one?

 

[A///]

Where can I buy one?

the siryn isn't up for retail sale yet since it came out a few days ago. I assume it'll cost between 10-20K for cpu alone. last gen "slow" altras top out at around $8k-10k for certain aibs in the arm space as whole workstation towers.

we're still a few years away from decent 1u arm servers on the used market, curious what the pricing on those would be. as it stands I could never buy a current gen arm server or workstation. it's far too expensive for a single system.

 

[coercitiv]

Where can I buy one?

Ampere Altra Developer Platform

COM-HPC platform based on Ampere Altra SoC (Neoverse N1), up to 80 Arm v8.2 64-bit cores up to 2.6GHz, up to 128GB DDR4 memory, 128GB NVMe storage, 64x PCIe Gen4 lanes (3 x16 available), 4x 10GbE (optional) and 1x GbE LAN, 4x USB 3.0/2.0 [rear I/O], 2x USB 3.0 [front panel], open source EDKII...

www.ipi.wiki

Have fun!

 

[H433x0n]

6+_8 for arrow lake is interesting. either intel cannot make an 8+16 part, which is likely the reason or their new 6+8 is powerful enough to blast past a 16 full fat cores processor for a total of 14 cores and 20 threads. this isn't the likely reason giving their past performance and behaviour.

if it ain't the 2nd reason they are so screwed for longer than most of us thought. I gave them benefit of the doubt. we were all so close to getting funk pops made of Pat "Guns Ablazing Gunslinger" Gelsinger made for the vast number of intel believers.seems a grand waste now.

I think it's simpler, MTL-S doesn't perform well enough to justify it's existence as a flagship desktop product. I don't think anybody is going to buy an expensive Z890 motherboard for a MTL i5 6+8 configuration, that just won't work.

It makes more sense to place it where it's best suited, which is mobile.

 

[A///]

I think it's simpler, MTL-S doesn't perform well enough to justify it's existence as a flagship desktop product. I don't think anybody is going to buy an expensive Z890 motherboard for a MTL i5 6+8 configuration, that just won't work.

It makes more sense to place it where it's best suited, which is mobile.

in that case neither does arrow. arrow here according to roach is severely cut down. wtf is going on @ hills borough?

 

[igor_kavinski]

wtf is going on @ hills borough?

Maybe ARL-S will be the first consumer platform to feature dual socket CPUs

Or dual compute chiplets.

 

[jpiniero]

I think it's simpler, MTL-S doesn't perform well enough to justify it's existence as a flagship desktop product.

That wasn't what was rumored though. They were going to use the Meteor 6+8 die as the low end and Arrow (presumably 8+16) as the high end.

 

[igor_kavinski]

6+_8 for arrow lake is interesting. either intel cannot make an 8+16 part, which is likely the reason or their new 6+8 is powerful enough to blast past a 16 full fat cores processor for a total of 14 cores and 20 threads.

It's possible, if the E-cores are somehow running on steroids, able to offer at least Zen 3 level performance.

 

[DrMrLordX]

incorrect. they've been regularly in the news. it doesn't get picked up because few are using arm power houses in their workflow outside the dc for the most part. they're not the only socketable system that mirrors an x86 setup for the most part. it's a real niche industry at the moe. their 192 tramples the 9654 while using less power. it's very much if you gotta ask you can't afford it type of deal. expanding the reason you don't see the reporting is no one really cares outside of big strides and product highlights. it's even rarer for anyone to get their hands on it vs other used or new hardware not meant for joe shmo consumers.

Feel free to link relevant stories in the appropriate threads, then. I should be subscribed to most of the ARM threads in the CPU forum, and I'd be interested in seeing what's going on with them.

Where can I buy one?

When Ampere came out with their 32c Altra, I wanted to get one in a 1P unit but nobody was selling them (that I could tell). The cited cost for board+CPU wasn't that bad. It would have been a fun toy at the very least.

 

[H433x0n]

in that case neither does arrow. arrow here according to roach is severely cut down. wtf is going on @ hills borough?

That wasn't what was rumored though. They were going to use the Meteor 6+8 die as the low end and Arrow (presumably 8+16) as the high end.

I don't see where ARL-S got cut down? Is there a new rumor/leak about that?

It seems like we went from this:
ARL-S 8P/16E (high end)
MTL-S 6P/8E (low end)

To this:
ARL-S 8P/16E (high end)
ARL-S 6P/8E (low end)