251–260 of 260 posts
Earlier quoted context omitted.
Itanium only failed because AMD was allowed to come up with AMD64, Intel would have managed to push Itanium no matter what, if there were no alternatives to a 64bit compatible x86 CPU.
But Itanium was not compatible with x86, it used emulation to run x86 software.
Earlier quoted context omitted.
Itanium only failed because AMD was allowed to come up with AMD64, Intel would have managed to push Itanium no matter what, if there were no alternatives to a 64bit compatible x86 CPU.
Itanium wasn't x86 compatible, it used the EPIC VLIW instruction set. It relied heavily on compiler optimization that never really materialized. I think it was called speculative precompilation or something like that. The Itanium suffered in two ways that had interplay with one another. The first is that it was very latency sensitive and non-deterministic fetches stalled it. The second was there often weren't enough…
Had it not happened, PC makers wouldn't have had any other alternative other than buy PCs with Windows / Itanium, no matter what.
Earlier quoted context omitted.
Itanium wasn't x86 compatible, it used the EPIC VLIW instruction set. It relied heavily on compiler optimization that never really materialized. I think it was called speculative precompilation or something like that. The Itanium suffered in two ways that had interplay with one another. The first is that it was very latency sensitive and non-deterministic fetches stalled it. The second was there often weren't enough…
I didn't said it was, only that AMD allowed an escape hatch. Had it not happened, PC makers wouldn't have had any other alternative other than buy PCs with Windows / Itanium, no matter what.
At Itanium's launch, an x86 Windows Server could use Physical Address Extension to support 128GBs of RAM. In an alt timeline where x86-64 never happened, we'd have likely seen PAE perk down to consumer level operating systems to support greater than 4GB of RAM. It was supported on all popular consumer x86 CPUs from Intel and AMD at the time.
The primary reasons we have the technologies we have today was wide availability and wide support. Itanium never achieved either. In a timeline without x86-64 there might have been room for IBM Power to compete with Xeon/Opteron/Itanium. The console wars would have still developed the underlying technologies used by Nvidia for it's ML products, and Intel would likely be devoting resource into making Itanium an ML powerhouse.
We'd be stuck with x86, ARM or Power as a desktop option.
Earlier quoted context omitted.
Well, lets not forget, AMD is AMD because they reverse-engineered Intel chips....
IBM didn't want to rely solely on Intel when introducing PCs so it forced Intel to share its arch with another manufacturer that turned out to be AMD. It's not like AMD stole it. Math coprocessor was in turn invented by AMD (Am9511, Am9512) and licensed by Intel (8231, 8232).
> Twenty-four 200 gigabit (Gb) Ethernet ports are integrated into every Intel Gaudi 3 accelerator WHAT‽ It's basically got the equivalent of a 24-port, 200-gigabit switch built into it. How does that make sense? Can you imaging stringing 24 Cat 8 cables between servers in a single rack? Wait: How do you even decide where those cables go? Do you buy 24 Gaudi 3 accelerators and run cables directly between every single…
One nice thing about this (and the new offerings from AMD) is that they will be using the "open accelerator module (OAM)" interface- which standardizes the connector that they use to put them on baseboards, similar to the SXM connections of Nvidia that use MegArray connectors to thier baseboards. With Nvidia, the SXM connection pinouts have always been held proprietary and confidential. For example, P100's and V100's…
Earlier quoted context omitted.
Hmm interesting, I was linked to an OCP dropbox with a version that did have the connector pinouts. Maybe something someone shouldn’t have posted then…
couldn't someone just buy one of those chinese sxm2 to pcie adapter boards and test continuity to get the pinouts? I have one that could take like 10 minutes
If you're willing to do the testing and post the pin out, I'll send it your way.
Would like to have it sent back when you're finished though, and I'd pay shipping both ways.
I do have the ground pins mapped out and could share that which should save some time.
Earlier quoted context omitted.
I look forward to seeing it. NVIDIA needs real competition for their own benefit if not the market as a whole. I want a richer ecosystem where Intel, AMD, NVIDIA and other players all join in with the winner being the consumer. From a selfish point of view I also want to do more home experimentation. LLMs are so new that you can make breakthroughs without a huge team but it really helps to have hardware to make it ea…
> I want a richer ecosystem where Intel, AMD, NVIDIA and other players all join in with the winner being the consumer. This is exactly the void I'm trying to fill.
Earlier quoted context omitted.
> I want a richer ecosystem where Intel, AMD, NVIDIA and other players all join in with the winner being the consumer. This is exactly the void I'm trying to fill.
I can't find your company, what are you trying to do to solve this?
Earlier quoted context omitted.
I would run simple llama.cpp batch jobs for 10 minutes when it would suddenly fail, and require a restart. Random VM_L2_PROTECTION_FAULT in dmesg, something having to do with doorbells. I did report this, never heard back from them.
Did you run on the blessed Ubuntu version with the blessed kernel version and the blessed driver version? As otherwise you really are in a development branch. If you can point me to a repro I'll add it to my todo list. You can probably tag me in the github issue if that's where you reported it.