Longhorn

21.2K posts

Longhorn

Longhorn

@never_released

Kernel/hypervisor engineer @awscloud EC2. Hobby @checkra1n. Mastodon: https://t.co/DsXP8PFgL0 Bluesky: https://t.co/dAOfFSSqY4

Berlin, Deutschland Beigetreten Haziran 2015
144 Folgt14.6K Follower
Longhorn
Longhorn@never_released·
to my knowledge CUDA 12.5 is the newest CUDA that works on Xavier
Longhorn tweet media
English
0
0
6
565
Longhorn
Longhorn@never_released·
@handleym99 @crramirezc The problem with Microsoft I think is more that they oscillated between the two instead of committing to one model, pissing off everybody
English
0
0
0
25
Maynard Handley 🟦
Maynard Handley 🟦@handleym99·
Well, that's if you believe that the basic model (varied "open" ecosystem, modular hardware) would have been the correct choice. The iPhone suggests that most of those are inessential to wrong. (You can claim Android is a counter-example, to which one could argue that the Android ecosystem basically waits for Apple to figure out new things, then copies them). etc etc That's my point - everyone can see what they want in the failure!
English
1
0
0
27
INIYSA
INIYSA@lafaiel·
If this claim is true, Qualcomm made an amazing leap in GPU RT acceleration. Or they just rendered it using CPU
INIYSA tweet media
English
23
2
129
49.8K
Longhorn retweetet
Eugenia Kuyda
Eugenia Kuyda@ekuyda·
we've reached peak journalism: Vanity Fair wanted to write a feature story about Dario but never got an interview so just asked Claude to make it up
Eugenia Kuyda tweet media
English
18
82
687
119.7K
Longhorn
Longhorn@never_released·
New adreno driver
Longhorn tweet media
English
3
2
40
2.7K
Longhorn
Longhorn@never_released·
nested virt with kernel-irqchip=off
English
0
1
5
1.6K
Longhorn
Longhorn@never_released·
@aspauldingcode backed by Hypervisor.framework (and ParavirtualizedGraphics.framework for the GPU device)
English
1
1
9
964
Longhorn
Longhorn@never_released·
Longhorn tweet media
ZXX
6
4
131
11.3K
Longhorn
Longhorn@never_released·
@j4567r I can put anything there :) as it's a virtual machine
English
0
0
5
916
Longhorn
Longhorn@never_released·
@FelixCLC_ That was really old PhysX versions and NV fixed it to something more reasonable back then Oh, and PhysX is open source nowadays
English
1
0
10
742
@fclc
@fclc@FelixCLC_·
have to wonder if a modern model might be able to binary patch its way to decent cpu physx performance?
LaurieWired@lauriewired

nah, old physX was bad and purposefully anti competitive. NVIDIA...let's say "strategically neglected" ageia's ancient codebase which used a ton of x87 floating point. If you know anything about x87...instructions are notoriously slow / scalar on x86 CPUs. They *could* have ported PhysX to SSE (and eventually AVX)...but that would mean it would run great on an ordinary x86 Intel/AMD cpu. By doing the "ooops, we left it on x87" they created a pipeline that would *only* look good on NVIDIA GPUs, and nothing else. Zero good reason to use x87 floating point for physics calculations in games. 80 bits of precision is ridiculous.

English
3
0
14
2.9K
Longhorn
Longhorn@never_released·
is it just me or DLSS 5 is reminiscent of the DLSS 1 story before they pivoted to making a TAA(U)?
English
10
0
33
5.6K
Longhorn
Longhorn@never_released·
but this still doesn't bring it near parity with the Windows version
English
0
0
2
696
Longhorn
Longhorn@never_released·
excel on mac has this and it's quite useful (KeyTips)
Longhorn tweet media
English
1
0
4
1.3K
Longhorn
Longhorn@never_released·
TAAs - with jitter added between frames so have more detail to accumulate - are theoretically sound. This is different...
English
0
0
8
1.2K

Entdecken

@handleym99 @crramirezc @lafaiel @vcprocles @aspauldingcode @j4567r @FelixCLC_ @elonmusk