Tony Wu Profile
Tony Wu

@Quarky93

Followers
216
Following
3K
Statuses
431

accelerating cryptography @Ingo_zk

Joined February 2022
Don't wanna be here? Send us removal request.
@Quarky93
Tony Wu
2 days
@__tinygrad__ The AVX512 instructions are supported on Zen 4 but doesn’t have advantage over AVX2. On Zen 5, AVX512 is “real”.
0
0
13
@Quarky93
Tony Wu
3 days
1
0
3
@Quarky93
Tony Wu
3 days
@ghost_motley Am I the only one that would like to see 24P cores + 16E cores instead?
2
0
3
@Quarky93
Tony Wu
3 days
@liquan_eth @drouyang @radi_cojbasic @Ingo_zk @Snarkify_ZKP @SuccinctLabs @RiscZero There's a layer below PTX called SASS. You can think of PTX as analogous to JVM bytecode, and SASS as analogous to x86 assembly. SASS is not officially documented by NVIDIA, but there are reverse engineering efforts.
0
4
4
@Quarky93
Tony Wu
3 days
@DistributedMarz It’s going to get much much faster
1
0
0
@Quarky93
Tony Wu
3 days
@thann199 This is on top of the speed up from stone to stwo (1000x) 😉
1
0
2
@Quarky93
Tony Wu
3 days
@AnnatarTheMaia @kai_fall Better compatibility though. Especially on Linux where there is a million different distros. Maybe people shipped statically linked binaries more, we would be so dependent on Docker these days.
1
0
1
@Quarky93
Tony Wu
3 days
@Starknet It will be real time
0
0
0
@Quarky93
Tony Wu
3 days
@CryptoKaleo Meme coins has destroyed crypto lol
0
0
0
@Quarky93
Tony Wu
3 days
@StockMKTNewz Just need to have Nvidia laptops and Nvidia phones now 😛
1
0
2
@Quarky93
Tony Wu
4 days
Le Chat
@MistralAI
Mistral AI
4 days
Le Chat is fast (1,100 tok/s for flash queries on an updated Mistral Large). Download it at or
0
0
2
@Quarky93
Tony Wu
4 days
@Uniswap Someone make this a physical machine 😂
0
0
0
@Quarky93
Tony Wu
4 days
I predict GPU accelerated STWO will get even more crazy fast within 6 months!
@Ingo_zk
Ingonyama
4 days
Introducing ICICLE-STWO A GPU-accelerated STWO prover that delivers up to 7x faster proofs than @StarkWareLtd's SIMD-optimized backend. Fully open-source and a drop-in replacement, ICICLE-STWO moves proving end-to-end to the GPU for unmatched efficiency. 🔗 Full breakdown:
Tweet media one
1
0
7
@Quarky93
Tony Wu
4 days
@rovercrc Looks like 1.5 data points to me lol
2
0
1
@Quarky93
Tony Wu
4 days
@0xsachi Depends on how they use AI. I have a feeling it’s going to be the same. Depends on the student not the tool.
0
0
1
@Quarky93
Tony Wu
5 days
@highyieldYT glorious, had the Radeon R9 Fury. Gen1 HBM with 4096-bit bus :D
2
0
7
@Quarky93
Tony Wu
7 days
@divine_economy Unfortunately it’s going to be retail that will sell right now
0
0
0
@Quarky93
Tony Wu
7 days
Funny how -20% drop in a day is a shock these days 😂💸 .
0
0
0
@Quarky93
Tony Wu
7 days
@sassal0x Hmmm what’s driving the current demand?
1
0
0