my123

24 ms·

421.
by my123 4y ago
> Intel's margins are strongest on servers Not anymore. Intel CCG is at 9.3b revenue/2.8b profit this quarter (30% operational margin). Intel DCG is at 6b revenue/1.7b profit this quarter (28% operational margin).
422.
by my123 4y ago
> your only option is to run all of FCP in Rosetta emulation. Apple has out of process plugins nowadays for that use case, through the XPC infrastructure. It's more 3rd-party SW that doesn't use XPC out of process plugins that
423.
by my123 4y ago
It however didn’t seem to make the move to Wayland… And unity8 (now Lomri) is in perpetual beta, and written in Qt.
424.
by my123 4y ago
> You can always emulate FP64 That's what's done by Intel for their newer consumer GPUs AFAIK, zero FP64 units present on die, but an emulation mode can be enabled to expose that support. (which is then done by software on the
425.
by my123 4y ago
Apple GPUs don't support FP64 at all - and Apple only provides Metal as a (supported and not deprecated) API. NVIDIA GPUs have much more ubiquitous programming models, from CUDA C++ to OpenMP (both C++ and Fortran), allowing them to be
426.
by my123 4y ago
Production readiness is a very hard thing to do. Not because you can make one unit you can make a bunch of them at acceptable yields.
427.
by my123 5y ago
Apple is building a web browser engine, Google is building a sandboxed application runtime - that happens to be a web browser engine superset.
428.
by my123 5y ago
Each contributor to Linux retains the copyright to their contributions. Google is a pretty big Linux contributor.
429.
by my123 5y ago
> what is the remaining work I would need to have done in order to get it working under A good chunk of the Linux community on phones didn't want to have to rely on binary drivers for non-technical reasons, and that's one of th
430.
by my123 5y ago
(disclaimer: I work at Qualcomm) > Mobile devices have no UEFI Since the Snapdragon 835, Android devices with Qualcomm processors use UEFI all the way to the kernel. However, custom UEFI binaries aren't part of the boot flow, with a
431.
by my123 5y ago
GPUs without graphics support are already a thing. But so far, NVIDIA is staying on the GPU term. For the H100, very minimalistic FF support was picked. Only two TPCs in both the SXM5 and PCIe H100 GPUs are graphics-capable (that is, they c
432.
by my123 5y ago
I actually asked and got told that it’s not a typo. But not more details beyond that so I can only speculate… And yeah, 192KB/128KB is what the HW behaves as by all accounts by measurements, making this even more odd…
433.
by my123 5y ago
That’s what Apple explicitly says. Why they don’t match with measurements is another matter… (and no, it’s not a typo) ECC? Needs hugepages? Or hardware erratum? I don’t know for sure. Also, at least from the VM perspective icache is report
434.
by my123 5y ago
> with their 128 kB data cache Bigger than that. ( https://github.com/apple/darwin-xnu/blob/2ff845c2e033bd0ff64... ) /* I-Cache, 256KB for Firestorm, 128KB for Icestorm, 6-way. / / D-Cache,
435.
by my123 5y ago
> at least at an end-user level I'd say that the most popular option at the end-user level today is a whole level of abstraction above: using Python w/ PyTorch (or less frequently, TensorFlow). Combined with CuPy and cuNumeric,
436.
by my123 5y ago
https://github.com/JuliaComputingOSS/llvm-cbe edit: you also have mrustc as a Rust to C compiler outright.
437.
by my123 5y ago
> And we have CUDA which is an attempt to, yes, get back to the C model CUDA allows different programming models, it's the point of having the PTX virtual machine as the exposed programming model. CUDA C++ is popular, but is far fro
438.
by my123 5y ago
> You need to get your software signed by Microsoft because it wants to be the arbiter of allowed software. Microsoft is deliberately not handling Authenticode certificates themselves in an attempt to stem this problem. Those certificate
439.
by my123 5y ago
And there are actual SC machines in the wild. For example, the Carmel cores on the Tegra Xavier processor provide SC guarantees.
440.
by my123 5y ago
That's not a cryptographic signature. Use codesign -s - path to binary to recompute the checksum.
441.
by my123 5y ago
It’s the EnterpriseG SKU.
442.
by my123 5y ago
All of TSX is present on Intel server processors. Just gone from client.
443.
by my123 5y ago
The VPU isn’t a VLIW at all, it’s an interesting RISC. Arm GPUs are also licensable by anybody, they were x86 phones with them back when those were still a thing.
444.
Public Preview for NXP I.mx 8M on Windows 10 IoT Enterprise(techcommunity.microsoft.com)
3 points by my123 5y ago|1 comments
445.
by my123 5y ago
It lived on for quite a while with a strong position in the networking market later on. But then instead of focusing on networking/infrastructure hardware like they could have, they went on a wild goose chase... to try to gain a place
446.
by my123 5y ago
Slowly dithered away...
447.
by my123 5y ago
> My experience with this, is moving away from the Nvidia kernel isn't practically feasible Oh yup... for just 4.14 (from the NV 4.9 kernel), took so much work to keep up... I eventually had a project to decouple nvgpu so that it co
448.
by my123 5y ago
> Support drops real fast for older boards, leaving one stuck on old kernels (Jetson Nano and its upstream family was "EOL'd" in terms of kernel upgrades a while ago, even though Nvidia will still gladly take your money fo
449.
by my123 5y ago
> I was talking about VK_NV_ray_tracing precisely, which is proprietary Vulkan extension Khronos does not promote an extension to KHR until multiple vendors ship support for it. VK_KHR_ray_tracing is a direct continuation of VK_NV_ray_tr
450.
by my123 5y ago
> GPU interconnects for direct GPU to GPU communication Note that nvlink is also available on the 3090, which does also have the same bus width as the A6000.

More ›