Intel Announces 48-Core Cascade-AP Multi-Chip Package

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Gideon

Golden Member
Nov 27, 2007
1,598
3,520
136
The only upside I see to this product is the socket and 12 memory channel motherboard. There are certainly some workflows that like that. But even that is only really any good once Intel manages to squeeze a nice multi-core Icelake, or rather Sapphire Rapid, into that.
 

Arkaign

Lifer
Oct 27, 2006
20,736
1,377
126
The only upside I see to this product is the socket and 12 memory channel motherboard. There are certainly some workflows that like that. But even that is only really any good once Intel manages to squeeze a nice multi-core Icelake, or rather Sapphire Rapid, into that.

Yeah I fully realize I'm not the customer for this kind of thing (really anything with gobtons of cores but slower individual core clock/perf), but it seems like good foundational progress that just needs some other of their tech progress to catch up with it. Mature sub-14nm, and I could see it doing way better. 1st attempts by Intel and even AMD can often be not without growing pains. And of course late pushes of older tech beyond comfort levels can yield similar disappointment in overall package. This thing seems like it may be a bit of a combo of these issues, eg : some new elements but a bit of growing pains, some older elements pushed beyond norms, and in need of some help.

Intel's first Pentium 60/66Mhz were expensive mostly pointless products compared to competing 486s with higher clocks. Pentium 75/90/100 started to make more sense.

AMDs K5 was pretty borderline at best. Virtually no OC, slower IPC. K6 otoh was banging.

Intel's first Slot 1 P2s were mediocre. Then Intel's P3-600 before the die shrink was damn near unstable from the jump, needing a .05v factory voltage hike lol.

Initial Athlon Slot A were pretty good tbh, but then the AMD chipsets were honestly trash fire. Was it 'Irongate'? Sound card issues, poor EIDE performance. To be fair, I believe that was pretty early for new chipsets exclusively designed, as previously they could just share industry standard and drop in with Intel sockets. Man, I miss those days lol. Goodbye Pmmx233, hello K6-3-333 :)

Intel again pushed things too far with the "1Ghz" race with initial batches being really at their limits or beyond right out of the box haha.

Intel's entire P4 Willamette launch was ass. Some admittedly cool innovations, but tied down with so much expense and compromises to Netburst. Anyone feel like paying a grand for 256MB of RDRAM? No? Me either.

Etc etc.

As we hoped with AMD to make things awesome again with Ryzen (they did). Let's hope Intel can poopcan 14nm soon. It's one thing to stagnate, it's another to have products being shoved out the door that clearly need better processes to live up to their design ethos.
 
  • Like
Reactions: french toast

DrMrLordX

Lifer
Apr 27, 2000
21,570
10,762
136
It's likely for density. There's an article that was saying the volume has been shifting from 8 socket systems to 4 as core counts balloon. Cascade Lake AP isn't the best example, but future efforts from both companies will steer towards 1/2 socket systems with 2x cores.

Does this product even improve density? Those sockets can't be small.
 

IntelUser2000

Elite Member
Oct 14, 2003
8,686
3,785
136
Does this product even improve density? Those sockets can't be small.

Have you seen the spacing between sockets for multi-socket boards? 4P to 2P will save a lot of space. It's not just that though. The -AP series replace Xeon Phi lineup. So it'll likely only support 1 DIMM per channel further saving space. Custom board designs by mega FLOPS farms will do even more. That's a big perf/density increase from the regular Xeons.

The Knights Mill Xeon Phi tops out at 320W so it isn't that big of an increase in TDP either, if anything.