JustSomeGuy1
About
- Banned
- Username
- JustSomeGuy1
- Joined
- Visits
- 60
- Last Active
- Roles
- member
- Points
- 1,172
- Badges
- 1
- Posts
- 330
Reactions
-
Inside Apple's fantastically fast new Mac Pro
hmurchison said:JSG1 - Thanks for the cleanup. I'm no Engineer admittedly but love to catch up every now and then. A few days ago I was looking some PCI-Sig videos about what they're doing with version 4, 5 and 6. The roadmap has incredible bandwidth increases. I totally misspoke and alluded to PCI limitations with NVME. It's not the limitation of PCI that I really wanted to say was that a NVME SSD on PCIE4 doing 5Gbps is going to differ in real world results depending on usage case.I for one and really glad Apple has a 6k Mac again. I'm old now...I remember when the IIFX was around $8k in a zero/zero config. For years Apple straddled the prosumer/workstation arena with PowerMac G4, G5 etc and many got used to these tweeners. The near future should be interesting as the Mac Pro internal design evolves with higher bandwidth solutions. It's never going to be cheap but if you make your living with that Mac you can often justify the purchase ...just hire a good accountant.JSG1 let me ask you about your personal feelings about the future of tech like Optane.No way to tell with Optane. Intel missed their target by a mile with the first release, and the next-gen Optane isn't a huge improvement. But... we've seen it before with Intel, where they keep iterating and eventually get to where they wanted to be. It's possible that Optane will yet live up to their promises. If so, then that's great, the world could really use a good SCM. If not, then it's a niche product, with some use for some people, but it won't move the whole market much.SCMs in general though... They're a revolution waiting to happen. The big questions are:1) Will they happen? Is there a tech coming that will really be able to replace both DRAM and flash? If I had to bet, I'd say yes, but I wouldn't feel very confident in that guess.2) Will the price ever hit mainstream markets? No clue about that one at all.Beyond SCM, there's in-memory compute. Not much like SCM, except for how it promises a major paradigm shift by changing how we think about memory. Interesting things happening there too, but it's too soon to try to predict where that's going. I think it's going to be big, though, unless it gets overtaken by something even bigger.One thing I will say: The desktops we use today are not really different from the desktops we used ten or twenty years ago, seen from 10k feet. You have a CPU, memory, I/O busses, ports, video, storage. Details have changed a LOT but all the parts are organized in roughly the same way. I think that that will NOT be true ten years from now. Big changes are coming.
-
Inside Apple's fantastically fast new Mac Pro
hmurchison said:
Yes and no. Don't expect a lot of innovation coming from Intel they're likely to skip next years PCI-Express 4.0 evolutionary step and refresh most of their Xeon class hardware with PCI-Express 5.0 which employs a lot of new technology chiefly a new CXL Interconnect . NVME SSD on PCI-Express 4 are already doing 5Gbps reads and very fast writes but many people feel like existing PCI-E limitations limit the benefits of that extra speed After PCI-Express 5 you're going to see 4-Level Pulse Amplitude Modulation (PAM4) which will not only get PCI-E 6.0 to 64 Gigatransfer but it'll be key in 100Gbps Ethernet for Data Centers and more.sdw2001 said:I can't get over what a monster this thing is. Apple's "pro" machines have always been more marketed to prosumers/power users rather than true workstation users. This machine changes everything.
I'm not up on PC workstation class machines, so a question for someone who is: Is there anything even close to this?The obvious question is who needs this bandwidth? AI, 8K and VR encoding, Machine Learning, Databases as always and even gaming. I think Apple has designed or workstation form factor that is going to be able to scale with the insane amount of bandwidth increases we have coming in just 4-5 years.You need to do a lot more reading, and you can start with the link you provided. You are confused about the relationship between CXL and PCIe5. PCI5 does not use any aspect of CXL. Rather, CXL leverages PCIe 5, with its major selling point being memory coherency between CPUs and attached coprocessors of various sorts (GPUs, FPGAs, NNPs, etc.). It's roughly similar to CCIX, and somewhat similar to GenZ (though CXL and CCIX use PCIe5 mechanicals, whereas GenZ doesn't).You are similarly confused about the state of play in NVMe SSDs. Not one person who understands the technology thinks that "existing PCI-E limitations limit the benefits of that extra speed". The limitation comes from the only PCIe4-capable controller that's on the market currently (the Phison PS5016-E16), which was a quick patch job on a previous controller to add PCIe4 compatability. Better controllers are coming soon, and they should be able to push read and write speeds up to ~7GB/sec... that is, if you're moving bulk data. For most people, random I/O is more important, as is latency, and the PCIe version doesn't make any difference for that- it's the flash and the controller.As for 100Gbps Ethernet, a single PCIe3 x16 slot is almost but not quite adequate to saturate the link. A PCIe4 x16 could handle a dual port card. By the time you get to PCIe5, a single x4 slot could handle a single port. PCIe6's bandwidth will obviously be welcome for anyone using 100Gbps Ethernet, but it's far from necessary.
-
Inside Apple's fantastically fast new Mac Pro
sdw2001 said:I can't get over what a monster this thing is. Apple's "pro" machines have always been more marketed to prosumers/power users rather than true workstation users. This machine changes everything.
I'm not up on PC workstation class machines, so a question for someone who is: Is there anything even close to this?Yes and no. You can get machines from HP, Dell, etc. with similar CPUs, RAM, and room for video cards. The afterburner card, no. That much Thunderbolt, no. Those video cards, no, though you can use NVidia cards that are either way better or somewhat worse, depending on what you do. Flash storage, yes, and you can do better than the Mac (FSVO better, again depending on use case). Slots... maybe not, I haven't checked.Unfortunately, this Mac *still* hasn't shipped, whereas EPYC 2 is now readily available, with Zen-2 based Threadripper coming soon. While some people will still be unwilling to look at AMD products, I doubt that that will last long, as the AMD chips are ridiculously superior to Intel's product line, and will remain so for at least a year, I expect.This Mac Pro will be a great workstation at a reasonably competitive price *for Intel-based workstations* (probably, but Intel pricing volatility and Apple's pricing stability may make Apple's pricing very unfavorable - time will tell). But cheaper Threadripper or EPYC-based workstations will wipe the floor with it, in most ways, and probably by the end of the year.On a separate topic, does anyone know if the TB3 ports on the PCIe card are somehow provided with full bandwidth? Or are they constrained by the bandwidth of the PCIe card?
-
Apple's Phil Schiller and Anand Shimpi tease details of A13 Bionic chip
morky said:I miss Anand. There is nothing like his A-series articles since he went to Apple.I disagree. Andrei F's articles at AnandTech now are extraordinary; there's nothing else out there even close, as far as I'm aware. His coverage of the A11 and A12 especially has been great.Contrast that with this Wired article, which was seriously terrible.
I don't find this surprising at all. They are clearly optimizing for power consumption, and they are hitting very strict power targets. I suspect it's no coincidence that they gained 20% on power dissipation, so they added 20% more capacity.wizard69 said:The other rather big surprise is the lack of a significant beefing up of Neural Engine. This makes e wonder if they will rely upon the ML extensions in ARM from now on. Their focus on ML techniques had me expecting a doubling of performance in the Neural Engine.For me the biggest surprise is that that while single-core CPU performance went up roughly 13-15%, multi-core performance went up maybe 19-20% (pending more benchmarks, especially Andrei's). There are two ways to explain this:1) The efficiency cores improved more than the performance cores, or2) Performance scaling is closer to linear than before, due to improvements in cache (coherency?), mesh/ring bus, power gating (not so likely), or...?#1 seems unlikely to explain most of this gain, as the improvements there would have to be really large to move the needle that much (though that's not impossible). #2 seems more likely. Any of those options would be good, and are particularly exciting as they all (except maybe the power gating explanation?) bode especially well for the future Apple laptop/desktop chip. In essence, this is suggesting that Apple will be able to deliver even better multicore results than was previously expected (based on the A12), and those expectations were already extremely high. They were high because it looks like Apple was able to achieve closer to linear scaling with their 8-core A12X than Intel was with, for example, the i9-9900. And now, scaling seems to have improved further with the A13 vs. the A12.It's not conclusive, but it's a really really good sign.[Edited for typo]
-
More power with less: Apple's A13 Bionic is faster and more power efficient
knowitall said:Really impressive especially the power savings and switching of active areas, Kudos to the design team.
The new Ad13 iMac will be a blast.
But Teslas FSD chip is currently 12 times faster (73 TOPS vs A13 6 TOPS?) which makes clear Tesla has a design team (only a few people I understood) that can easily match Apples.
So I expect much room for improvement for the A14 and its desktop version the Ad14 next year.
Exciting times, it must be difficult working at Intel now.
Edit: note the TOPS (not TFLOPS)Tesla's chip is much faster than Apple's at running, say, resnet 50. But it's MUCH slower than Apple's at running any normal app.You're failing to distinguish between different types of TOPS. Tesla's achievement is notable but not nearly in the same category as Apple's. In principle, it's not all that hard to add more TOPS, if you're talking about tensor/matrix ops, vectors, NNI, GPU, etc. Those ops (and OPS) are all easily parallelized. Further, you fail to recognize the hard limits placed by the power and cooling budgets for each chip. Tesla is entirely focused on video image recognition, so they need massive NN processing. They have the power budget of a car - not unlimited, by a long shot, but still... the battery in a Tesla is a little bigger than the battery in an iPhone! Whereas Apple is building a much more general-purpose chip, and specifically one with extraordinary traditional integer (and FP) OPS. That's a MUCH MUCH harder problem to solve, as it's extremely difficult to extract parallelism from conventional software (that is, pretty much every app that isn't doing AR or a few other very specific things).So far, Apple has in the last couple of years kicked *everyone's* ass at that, inside their domain (low-power chips). Nobody even comes close. And if you look carefully at what they've done, you can build a fairly convincing case that they've already built every part necessary to beat Intel at their own game (fast high-power multicore chips), they just don't want to sell those yet.The biggest open question is this: Can Apple build a ring/mesh/whatever connecting 8-16 high-performance cores in a reasonable power budget? As I've written previously about the A12X, they've *already done this*. So they can, right now today, build something competitive with Intel's best mainstream desktop CPU (the 8-core i9-9900). Whether or not they can actually bet it will depend on whether or not they can clock up. And we know more about that than we did a few months ago, as we can see AMD pushing the same process to around 4.3 GHz tops, about 4.1 comfortably. We still don't know if the A13's pipeline is long enough to sustain this sort of speed, or how easy it would be for Apple to change it enough, but the performance crown there seems easily within their grasp.Going to more cores is the biggest question mark. The ring or mesh good enough to handle 8 cores really well may not be enough to handle 12 or 16 cores. But the only machines where that would matter is the iMac Pro and the desktop Mac Pro. And I don't think anyone expects those to transition to ARM as early as the laptops.
