New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Intel's 288 core whopper
Only E cores, but that is quite a few... 
Comments
Monster! looking for a $3.2/y deal
will it run crysis?!
That shit is probably going to cost like $30k 🤣
We need YABS !!!
Only E-cores makes this less appealing tbh. At least for SMEs. If you're trying to run a hyperscaler with low-use VMs then ... sure.
Think web servers. Tons of requests, but each request is very likely I/O or network-bound, not CPU bound. Having plentiful parallelism is more helpful than maximum CPU power.
It features AMX? What a monstor.
Point. I just don't build the kind of apps that would require this level of parallelism in web servers so it didn't occur to me. Most websites I use or make don't benefit as much, especially if storage/asset serving is offloaded to someone else.
In theory, yes. Practically though (a) IO (incl. network) needs CPU as well, and (b) there's a limit to how many IO tasks a core can handle. And (c) many data need some processing too.
ai though loves and needs many cores and that, I guess, is what that processor really - at least mainly - aims at; it has AMX for a reason ...
Of course I may be wrong but the way I see it is that Nvidia is the "performance" ai player, AMD is the (general computing) performance leader, so intel tries to get a bite of the cake by offering a "cheap" alternative.
Btw, again re "VMs and lots of them": drives are likely to be a bottleneck.
All in all I'm neither impressed by nor interested in that processor (except some design challenges).
Staggering amount of cores. Even for E class cores this is massive. I wonder how it would go in compare to threadripper in terms of price per watt.
I have some xeon phi processors (10 year old tech) in the basement which have also an enormous amount of cores 60 to 72 cores depending of the version, only intel stopped suddenly development and support for this cool tech. I never understood their logic behind this. I think arround 2015 was the start of their downfall, i think they were betting on too many horses around that time. And some financial manager used the cost cutting mechanism (axing whole teams/productlines) to keep their margins healthy.
That's a lot, but it's probably trash for heavy workloads with a horrible IPC since all the parts that made a CPU fast will have been reduced (register file, FMA units, per-core caches, branch predictor buffer, uop cache, OOO capabilities, etc.). All it'll be good for is running crap in parallel without as much context switching overhead (good for some kind of databases and stuffing lots of cheap VMs on a single node, I guess). More cores at the expense of everything else is rarely better.
yabs!
They were just never any good, sitting in an uncomfortable position between CPUs and GPUs. They didn't have the fast latency-minimizing sequential performance of a CPU, but they also lacked the raw parallel arithmetic capabilities of a GPU.
Now, if we're talking about promising technology that Intel killed, I think we can all agree that Optane shouldn't have been discontinued... It was an absolute god for high-speed caching without any wear.
I had many problems with optane on my hp laptop. That shows again that Intel bet on too many horses. They wanted to compete in so many fields SSD technology, HPC, graphical processors, fpga's, cpus, memory, mobile devices.... etc.etc.... They forgot how to grow in their own field; cpu's