sliken

32 ms·

511.
by sliken 3y ago
Dunno, the vast majority of phones, tablets, laptops, and desktops have had 64 or 128 bit wide memory interfaces for decades. Apple has a high profit market and a vertical market, unlike the competition. Apple skipped GPUs and did it thei
512.
by sliken 3y ago
Wow, 12 hour video, I do wish a small summary was posted. Even just a sentence or two, like your post, thanks. Sounds like the OpenGL work can mostly be reused for improving the Vulkan driver and sounds quite promising.
513.
by sliken 3y ago
Wow, excellent. I'm trying to justify replacing a 2015 desktop with either a mac studio or a ryzen/zen4 desktop. I'm not a gamer, but do occasionally fire up steam to play something old like Xcom 2 or Majesty. Knowing that
514.
by sliken 3y ago
What's old is new again. This is basically an updated arm version of the itanium based SGI Altix. Keep in mind unified does not mean uniform, the ram is distributed across all the GPUs.
515.
by sliken 3y ago
Proton is a wrapper/tweaks to wine, right? Thus x86-64 windows games can run on x86-64 linux. Thus the steam deck and many happy linux gamers. Apparently it's good enough to run a large majority of steam games, I believe I reme
516.
by sliken 3y ago
HP Dev one works well, kudos to System76 for the desktop env. It has the worst LCD panel I've seen in many years. It's not the resolution just poor contrast and poor color accuracy. It pains me to see it. They went on sale recen
517.
by sliken 3y ago
Apple makes good hardware and makes good engineering decisions. Sure other laptop makers have low end and high end options. But frustratingly often match apple on 50-75% of the features. Sure some have nice still aluminum bodies. Some h
518.
by sliken 3y ago
Soldered in memory is a key part of why apple hardware is lower power, smaller, and higher memory bandwidth than the x86-64 counterparts. Sure 2 dimms on a mac mini, or mba wouldn't hugely change the form factor. However 4 dimms, 8 d
519.
by sliken 3y ago
Sure, but that's a real number, not peak. The top ryzen (7950x) or top intel (i9-13k) manage only 83GB/sec peak, and something in around 60GB/sec real. Generally the arm memory model is looser than the x86-64, which allows
520.
by sliken 3y ago
I believe the chiplet links are pretty generous, they handle L3 <-> chiplet traffice and chiplet <-> chiplet traffic. On the smaller configs they use 2 links per chiplet. Thinks like McCalpin do not seem to show much difference
521.
by sliken 3y ago
> The problem is that for AMD it's not a good business decision. The PS5 and XboxX have an AMD APU (CPU+IGPU) with a wide memory interface. Seems like a fine decision. What surprises me is they haven't brought it to low/
522.
by sliken 3y ago
Sure, but what if a normal C code needs more bandwidth? Or if a GPU code needs more than 12-16GB of memory (normal cards) or 24GB (if you get a 4090)? What I like about the apple approach is that low end laptops/desktops get 100GB&#x2F
523.
by sliken 3y ago
Likely better on any workloads that are GPU heavy and fit in 24GB of vram. Two data points from Apple: a) M2 Ultra 24 cores, 128GB ram, 2TB storage = $5,200 b) M2 ultra 24 cores, 192GB ram, 4GB storage = $6,600 Likely 1/4th th
524.
by sliken 3y ago
192GB ram does seem like enough for a large fraction of the potential market, especially @ 800GB/sec which makes it much more usable than similar amounts of ram on Intel/AMD desktops at 1/8th the bandwidth.
525.
by sliken 3y ago
There does seem to be a general misunderstanding of what Peer-Reviewed means among the public. At my kids highschool there was an AIM program for harder classes for the academically inclined. A parent's kid didn't get in, they d
526.
by sliken 3y ago
My understanding is that the vast majority of them end up in products. Embedded in all kinds of systems. IoT, assembly lines, signs, lobby displays, interfacing with all kinds of dumb systems to get them an internet interface. $5 for a P
527.
by sliken 3y ago
It's never been that the supply was particularly low, just that the vast majority of them were going to large orders to be included in products. Much like the GPU vendors were selling directly to miners.
528.
by sliken 3y ago
Significantly more = 6-8 times the compiling performance. I was quite impressed as how fast the RK3588 is.
529.
by sliken 3y ago
I think you have it backwards. North America has Moose[0] (called elk in Eurasia) and elk [1]. I've seen them both in person. [0] https://en.wikipedia.org/wiki/Moose [1] https://en.wikipedia.org&#x
530.
by sliken 3y ago
That's like calling Tesla a battery maker. You can't buy a bare battery from Tesla or a CPU from Apple.
531.
by sliken 3y ago
Typically: 1. Low latency network, 1-2us. Most servers can't ping their local switch that quickly, let alone the most distant switch for 1M nodes 2. High bandwidth network, at least 200gbit 3. A parallel filesystem 4. Very f
532.
by sliken 3y ago
Get some FRS radio (walky talky type) in a bubble pack, maybe get a few extras for neighbors, and befriend your local ham. Talk to neighbors about emergency plans so you can coordinate helping the needy, tracking water/food deliveries
533.
by sliken 3y ago
Yes, but AMD GPUs haven't (so far) pulled off any of the triumphs that AMD CPUs have. The ones that spring to mind: * Intel tried to reserve 64 bit for servers/Itanium, AMD brought out x86-64 across their line * Intel was p
534.
by sliken 3y ago
What we need is more competition. Nvidia has plenty of substantially faster cards they are shipping, they are just minimizing performance delivered and maximizing cost to the consumer. 4070 is comparable in bandwidth to the 3060 Ti and wou
535.
by sliken 3y ago
Just seems like a money grab. The 4060 Ti has half the memory width (128 bit vs the old 256 bit wide) of the 3060 Ti, not much more than half the memory bandwidth. The old 3060 Ti has 448GB/sec and the new 4060 Ti has 288GB/sec.
536.
by sliken 3y ago
Sure, but Apple's doing pretty well there. Think of the apple silicon as a chiplet. Said chiplet has a 400GB/sec memory bus, 8 performance cores, 4 efficiency cores, and 38 gpu cores (not to mention accel for video encoding, mat
537.
by sliken 3y ago
My memory is that the pentium pro had a common memory config of 256MB, way below what modern installers are going to expect. I'm sure you can get linux to install on 256MB, but I doubt it's going to work on the current RHEL/
538.
by sliken 3y ago
Can use 16k pages. M1 also can use 4k pages.
539.
by sliken 3y ago
My kid got her android phone stolen from a gym. Put up a message on the phone saying "this phone is being tracked, reward, call XXX-XXX-XXXX." An hour later a grandfather called saying his grandkid would be returning the phone in
540.
by sliken 3y ago
I believe alder lake and newer intel desktop chips support ECC, if you get the right motherboard/chipset. Desktop ryzens support ECC for several generations, but ECC support varies by chipset. AMD forums often discuss it. Some Ryzen

More ›