Sailesh Kumar
@saileshtalks
Followers
2,053
Following
23
Media
3
Statuses
188
Design industry leading semiconductor solutions and help advance the industry. CEO at @bayasystems . ex- Intel Fellow and Founder at NetSpeed Systems.
Los Altos, CA
Joined November 2023
Don't wanna be here?
Send us removal request.
Explore trending content on Musk Viewer
Tuchel
• 164446 Tweets
#FGO
• 113600 Tweets
#sbhawks
• 48780 Tweets
カズラドロップ
• 47629 Tweets
バーニス
• 38359 Tweets
#deprem
• 36482 Tweets
#baystars
• 34164 Tweets
FY RECAP BLANK SS2EP4
• 31218 Tweets
#BlankReactSS2Ep4
• 28528 Tweets
ZETA
• 23817 Tweets
ソフトバンク
• 19958 Tweets
ジェロニモ
• 19709 Tweets
ホークス
• 17286 Tweets
スタメン
• 14523 Tweets
ジャイアンツ
• 14382 Tweets
カノウさん
• 12440 Tweets
ベイスターズ
• 12014 Tweets
日本シリーズ
• 10773 Tweets
Pinned Tweet
One of the most exciting days of my career.
@bayasystems
exiting stealth and launching at DAC. Thanks to BoD
@jimkxa
and
@stanreiss
, and
@intel
and
@PGelsinger
for support and investment, and numerous customers and partners.
My blog on Baya's inception:
32
31
102
Silicon Valley is going back to its roots. Over the next decade, huge value will be created in silicon. Those who lead in silicon design and manufacturing will win big. Pendulum has swung again. These cyclical trends last over decade(s). PS - this is not April fools joke.
32
86
1K
We are already in the era of multi PB/s bi-section fabric bandwidth within package.
If you are playing the high performance computing or AI silicon game and not currently designing your chiplets and interconnects for these levels, you have already lost.
4
13
331
Congestion Control (CC) is crucial for optimal (not over designed) fabrics. Even more so in larger faster systems. Most CC designs look dumb, weak or a mess. We are advancing the state of the art in CC. DM me all congestion control and fabric geeks if interested in a deep dive on
4
5
46
@jimkxa
@realGeorgeHotz
Less is more.
Less code, less gate, less opcode, less feature.
True in software, hardware, protocol, design, engineering.
It's not just desirable, it's a survival strategy.
7
6
36
Also a firm believer that
@intel
's future is brighter with
@PGelsinger
.
@jimkxa
had similar plans to fix
@intel
but now water under the bridge. Fortunate to have experienced both legends.
Another wishful thinking: Pat + Jim >>> Pat, Jim.
Embracing change,
@intel
is on a voyage to a brighter future. Kudos to
@PGelsinger
for steering the ship.
@jimkxa
was pursuing this and more 6 years ago, now wishful thinking. But as they say, better late than never! Future’s bright,
@intel
’s ready! 😎
0
2
7
2
1
14
@jimkxa
100% agree on the need of an truly open fabric protocol. Most current standards (open or closed) are driven by special interests of large corporations.
Internet revolution started after IP/Ethernet arrived as a true open standard. Computing and semiconductors need the same.
0
1
19
@EnerTuition
@bayasystems
@Arm
@intel
@AMD
@nvidia
@tenstorrent
@jimkxa
@RajaXg
ARM -> RISC-V exodus does not impact AMBA adoption including by RISC-V.
0
1
17
@Om_Prakash_Hari
@nvidia
@AMD
@intel
@tomshardware
BTW
@RajaXg
built the Ponte Vecchio GPU at
@intel
which was ahead of even Blackwell in many technological aspects. Hope Intel keeps advancing on that platform.
2
1
14
Jerome Powell: Jensen and Taylor are the top two drivers of inflation now. Higher rates for longer.
Jensen Huang: The more (chips) you buy the more you save.
Taylor Swift: You did a number on me, but honestly, baby, who's counting?
0
1
16
@PatrickMoorhead
During Network Processor mania couple decades ago the ASIC- and GP- like arch duelled. ASIC was always deemed efficient but GP won in the end. The reason: 1. workloads keep changing and GP adapts faster. 2. key bottlenecks are hw accelerated so GP is not far in efficiency.
1
0
9
@AlteraFPGA_
Congratulations to team
@AlteraFPGA_
and
@SandraLRivera
for leading and bringing back the iconic brand.
@bayasystems
cheers and roots for a bright future.
0
3
10
@PGelsinger
@intel
@CommerceGov
Congratulation
@PGelsinger
and team
@intel
.
The best days of Intel are ahead under your leadership.
0
0
6
@RajaXg
Five pillars: Compute, Memory, Communication, Power, and last but not the least Manufacturing
Compute margins are way higher than the rest due to software. Other 4 are software neutral.
0
1
4
@AMD
Revenue $5.5B, +2% yoy, Market cap 250B
@intel
Revenue $12.7B, +9% yoy, Market cap 125B
Wild and insane. What am I missing?
🤔 Something to ponder:
@intel
posts a $12.7B revenue in Q1 2024, up 9% YOY, yet their market cap sits at $125B. Meanwhile,
@AMD
, with $5.5B in revenue, enjoys a market cap of $250B. 📊 Twice the cap with half the revenue? What's driving this valuation disparity?
#TechInvesting
1
1
8
3
2
8
On May 23, 2024 NVDA will be the most valuable company in the world!
05-23-24
0
3
8
@EnerTuition
@bayasystems
@Arm
@intel
@AMD
@nvidia
@tenstorrent
@jimkxa
@RajaXg
Mostly agree.
Ethernet will emerge as the dominant scale out most likely.
Scale up perhaps a toss between AMBA, NV, Infinity. AMBA being open may surpass others in the end, just like it did eventually (?) in the ISA wars.
1
0
8
Unfortunately this is true for many CEOs and senior management. The results are clear. Tech companies run by techies excel more often than those run by bureaucrats, accountants and finance people.
The CEO of an aircraft company should know how to design aircraft, not spreadsheets
10K
29K
176K
2
0
5
@IanCutress
Chip size has not grown much over last 4 decades while transistor count has exponentially. The machinery of manufacturing, testing, packaging, and deployment is apparently built for fixed 2D dimensions. I wonder if the next frontier of super scaling will lead to bigger chips in
2
0
6
@HPC_Guru
@intel
Every chip design company’s goal should be to split manufacturing 50:50 between
@intel
and
@TWSemicon
before 2030. This is the only path for them to steady growth and avoid any nasty surprise. Their BoD must ensure it.
@nvidia
@Broadcom
@Qualcomm
@AMD
call
@PGelsinger
3
5
6
@tomshardware
In one word: insanity.
A healthy competition between the
#USA
and
#China
is gonna be the greatest accelerator for technology and for innovation. It's sad that walls being erected to stop it.
1
0
6
@RajaXg
New tech creates more jobs than it displaces. AI will be no different. Your observation is apt and will happen 100%.
More jobs will be created to feed the beast and to ride it.
Telecom (Bell Labs) era
Computing (IBM, DEC) era
Internet (Cisco) era
Personal Computing
2
2
6
@magicsilicon
@IEEESpectrum
@intel
3D stacking “improves the latency between compute and memory by shortening the hops, while at the same time enabling a larger cache”— Pushkar Ranade
This is the future.
@bayasystems
Cache Studio and Chiplet Studio are coming soon!
0
3
6
0
1
3
@InternetH0F
To be successful in anything Either you get lucky or you work hard or some combination.
Waiting to get lucky is clearly a bad strategy.
1
1
4
@EnerTuition
@bayasystems
@Arm
@intel
@AMD
@nvidia
@tenstorrent
@jimkxa
@RajaXg
"who is going to bat for AMBA?"
Rest of the world.
2
2
4
@Arm
@Arm
tech and strategic edge far surpasses all compute silicon companies and almost all
@Arm
customers. Long term they will win over all, especially the margins. Great future ahead.
0
1
5
@HPC_Guru
@intel
@TWSemicon
@nvidia
@Broadcom
@Qualcomm
@AMD
@PGelsinger
Yes they should be. Strategically speaking, I think the right balance may be 50:50 between
@intel
and (
@TWSemicon
+
@Samsung
) for many reasons including splitting capacity evenly across Pacific Ocean.
0
1
4
@bayasystems
@jimkxa
Yes.
And keep in mind that as things scale bandwidth grows faster than compute.
0
1
5
@PGelsinger
@intel
Congratulations
@PGelsinger
and
@intel
. The best days of Intel are ahead under your leadership.
1
0
5
@Om_Prakash_Hari
@nvidia
@AMD
@intel
@tomshardware
As
#GPU
compute becomes mainstream and software libraries keep advancing to leverage its higher raw compute per watt, per sqmm while providing growing programming abstraction, no top compute company can leave this race. It got to be a 3 horse race
@nvidia
@AMD
@intel
.
0
0
5
0
1
5
@alifarhat79
Here we are at all time highs and they are talking about cutting rates with inflation still raging and Congress spending like there's no tomorrow. The show is unbelievable.
0
0
3
@PGelsinger
@StabilityAI
@intel
AI semiconductor race has just begun and it's going to be a marathon not a sprint, and a very rewarding one, for many players not just one.
0
3
4
@intelnews
Congratulations to team
@AlteraFPGA_
and
@PGelsinger
and
@SandraLRivera
for bringing back the iconic brand.
At
@bayasystems
we cheer and root for a bright future ahead.
0
2
4
0
0
4
@roninhahn
@bayasystems
@jimkxa
@nvidia
They do not need alternate reality to justify a $1T valuation. They will probably hit $200B revenue and $100B profit within 2-3 years.
They will, to hit $10T, which many analysts are calling.
1
0
4
@Arm
@intel
@AMD
@Qualcomm
This is the fruit of owning main IP, fostering ecosystem + open/free standards (but managed).
#AI
barely starting.
@nvidia
like Xeon/Sun in heyday or RIMM/Blackberry when Mobile started. Explosive growth, high margin, but closed, lil HW IP ecosystem. Who will be
@Arm
of AI? (2/2)
0
2
4
0
0
3
@zerohedge
Higher for longer. Thanks to Nvidia and Taylor.
Jerome Powell: Jensen and Taylor are the top two drivers of inflation now. Higher rates for longer.
Jensen Huang: The more (chips) you buy the more you save.
Taylor Swift: You did a number on me, but honestly, baby, who's counting?
0
1
16
0
1
3
@bayasystems
@Chip_Insider
Gradually, then Suddenly. Those who don't see the change lose. Those who plan ahead win.
0
2
3
@RajaXg
Looks amazing. Will be fun to see how it scales, whether into bigger head or many heads...
0
0
3
@FelixCLC_
Cache and memory stack’s role in
#cpu
performance is often under appreciated. It’s all about feeding the
#cpu
pipeline at the lowest latency and highest throughput. Multi-core cluster performance is even more dependent on cache hierarchy’s nimble data movement both downwards and
1
0
2
@PatrickMoorhead
I will bet my money on General Purpose (GP) and programmable arch. (not necessarily GPU) vs ASIC.
0
0
3
@jdalrymple
Apple + ChatGPT would have been awesome
But looks like we will get Apple + GeminiCrap
0
0
3
@bayasystems
@tomshardware
@jimkxa
@AMD
@Tesla
@Apple
@intel
is now following his plan and will soon reap the rewards. 6 years of
#FAFO
.
0
0
3
@jimkxa
TCP/IP + Ethernet runs almost all off-chip communication in the world. There two are perhaps the most important technologies right besides Turing type computers (all programmable computers)
AMBA runs majority of on-chip communication.
0
0
3
@jimkxa
Depends on where we start redesigning physics from:
Caveman era -> Mythological physics era -> Newtonian era -> Relativity era -> Quantum era -> ?/AI/mythological/magic/unreal/simulation era
Starting from Caveman era will be fun. I strongly feel the redesign will follow the
1
1
3
@DrJimFan
One more driver of inflation. Higher for longer.
Jerome Powell: Jensen and Taylor are the top two drivers of inflation now. Higher rates for longer.
Jensen Huang: The more (chips) you buy the more you save.
Taylor Swift: You did a number on me, but honestly, baby, who's counting?
0
1
16
0
1
2
@jonmasters
@Arm
It does not have to do anything with
@arm
architecture, RISC-V or x86. It has everything to do with the scale of the chip, memory, cache, and fabric architecture.
So Kudos to Apple Silicon.
2
0
3
@IanCutress
Are they aiming to be # 1 or content with # 2? Intel historically shines when they are # 1 not # 2.
1
0
3
@shreesh_riscv
@intel
@PGelsinger
@jimkxa
Wishful thinking for sure but you never know.
@PGelsinger
@jimkxa
0
0
2
@texasrunnerDFW
We have one more driver of inflation now, Jensen Huang and Nvidia.
Jerome Powell: Jensen and Taylor are the top two drivers of inflation now. Higher rates for longer.
Jensen Huang: The more (chips) you buy the more you save.
Taylor Swift: You did a number on me, but honestly, baby, who's counting?
0
1
16
0
0
1
@ultrawide219
Intels biggest setback in GPU computing was
@RajaXg
departure who was the mastermind. Now they have no credible leadership.
1
0
2
0
0
2
@jimkxa
Compute always gets exponentially cheaper with time and competition will take care of the margins.
0
0
2
@magicsilicon
I wonder how pre-birth training compares to training for the rest of the life. I would assume most of pre-birth smarts is imprinted as natural reflexes in the entire body muscles, while brain is still mostly blank slate.
0
0
2
@Mojo_flyin
@bayasystems
@Arm
@intel
@AMD
@nvidia
@tenstorrent
@jimkxa
@RajaXg
Based on # of instances.
Revenue is difficult to quantify as a lot of development and deployment will be in-house.
0
0
2
@magicsilicon
2 weeks late :-)
On May 23, 2024 NVDA will be the most valuable company in the world!
05-23-24
0
3
8
0
0
2
0
0
2
1
0
2
@RajaXg
Interconnect bandwidth needs grows faster as compute grows. How to suppress bandwidth needs? More and bigger caches, all types of them. That’s a factor.
0
0
2
@MarioNawfal
You missed Fed, Nvidia and Taylor.
Jerome Powell: Jensen and Taylor are the top two drivers of inflation now. Higher rates for longer.
Jensen Huang: The more (chips) you buy the more you save.
Taylor Swift: You did a number on me, but honestly, baby, who's counting?
0
1
16
0
1
2
@kaizen_hower
Not necessarily. Moore's law can do its thing but addition of 3D scaling would be Moore's law on steroids.
1
0
2
1
0
2
Jerome Powell: Jensen and Taylor are the top two drivers of inflation now. Higher rates for longer.
Jensen Huang: The more (chips) you buy the more you save.
Taylor Swift: You did a number on me, but honestly, baby, who's counting?
0
1
16
0
0
1