Passionate about computer HW and SW design & architecture. Walking the Way of the Fool.

São Paulo, Brasil
This is the tweet that coined 'Silicon Gang' after our dear Ashraf retweeted it in hashtag format. The memories! #tdpWars #muhSkylake #amdomination #SiliconGang ♥️
Finding about your Twitter account and the silicon gang has been one of the most fun and remarkable things from 2018 to me! That and the fact Twitter could actually prove useful. :-)
5
1
35
Apple's A17 Pro perf cores = 10% faster (on 3nm, haha) Congrats on becoming the Intel of the last decade, Apple. 3rd subpar SoC in a row. The brain drain is real. #SiliconGang
23
26
236
34,803
It’s embarrassingly sad to watch this kind of intellectual dishonesty from Intel. Yeah, Raptor Lake is your latest uarch, “re-launched” this year with zero changes vs last year’s RPL except artificial segmentation through APO — and that one was but a less handicapped ADL. #SiliconGang intel.com/content/www/us/en/…
25
11
156
23,103
NVIDIA's RTX 4090 (Founders and all custom designs) is the single most horrible, hideous, and ominous piece of client hardware I have ever seen or imagined. It's uglier than even NVCP. 🤢🤮😷 I sincerely hope RDNA3 kicks NVIDIA to the curb this generation. #SiliconGang
15
16
136
Why is Apple still so hostile and neglectful towards PC gaming? No Vulkan, no apparent relevant DevRel, mostly no AAA titles. All these M1 SoCs with huge GPU area that you can't use for the biggest graphics use case and ~40B USD market. #SiliconGang
30
11
132
Absolutely impressive. Intel — while remaining at a + version of the infinitely maligned 10nm node — has made a huge efficiency job on ADL -> RPL, whereas AMD went firehouse even with N7 -> N5 on Ryzen 7000 (+~60% power for +~45% perf). Congrats to Intel silicon gangsters! ❤️💋
18
12
123
Windows 11 should drop 32-bit support for good.
17
7
93
AMD has delivered one of the worst gen-on-gen client uarchs & products in a long time with Zen5%. It’s Joever for them in desktops, with Zen6 being focused on packaging and stuff. Arrow Lake with Lion Cove and “Chadmont” will clearly pull ahead, and the gap will only increase by 2025-2026. It’s past time AMD should have increased core counts, but they’re the new old Intel with plenty of complacency. #SiliconGang
33
10
95
23,264
Intel: “here’s some Alder Lake that can boot Windows.” AMD: “here’s some v-stacked upcoming Ryzen that performs ~15% better than our current ones; watch it game.” There is a world of difference in execution, confidence, and ability to generate excitement. Sad about Intel.
5
8
89
Apple seems to be going downhill fast like the Intel of Krzanich if you look at the proper places. Rehashed CPUs uarchs for 3-4 gens, subpar YoY perf increases vs AMD/Intel/ARM, Android/QC has already caught up with A17 Pro on MT (and almost on efficiency?) even with a full node disadvantage, ever increasing power limits for iPhone SOCs due to stagnant IPC, extremely buggy iOS releases until version x.3.x at least, still no nested virtualization in MacOS for Apple Silicon, talent exodus is real (and results speak for themselves)… Cook’s successor may seal the deal… #SiliconGang
16
5
90
18,014
I’m glad to say your [least] favorite AMD fanboy & shill seems to be without symptoms and ready to storm Twitter again. Thanks for all the kindness during my recovery! Love you all — you’re amazing! ❤️🥰 Now, what have I missed in the past few weeks, #SiliconGang?
25
1
83
Apple's A16 = 3rd mediocre, dud SoC in a row, with no meaningful improvements to the CPU cores at all. Have all motivated CPU designers left Apple or what? Real question. At this rate, Apple Silicon for PCs will eat the dust really hard and soon. #SiliconGang
28
7
74
Replying to @reinaldoazevedo
Qual a saída? Entregar o país, balbuciando sobre a solução ideal no Estado Democrático de Direito enquanto o outro lado fala por meio de artilharia?
4
1
75
Zen5 desktop bloodbath. AMD will go “KS” levels of madness. #SiliconGang
10
5
74
13,537
My 13900K situation: - A couple months ago, had to go from SVID Typical to FailSafe and increase LLC from 3 to 4. - Since earlier this week, I have had to increase LLC to 5 and reduce multipliers to 55 and 41. Otherwise, League of Legends and vkg.sys would crash the PC with a BSOD or freeze every other game. Sigh. It’s dying. #SiliconGang
15
5
69
13,103
I shall shill hard. 😍❤️ #SiliconGang
7
6
74
Omg, on Z690, ASUS has added an anime 'Se7en' theme to the BIOS.
14
9
67
What madness is this? I knew Blizzard’s engine was CPU bottlenecked, but... My Heroes of the Storm FPS (Ultra/Max *everything*) with the RTX 2070 went 3900X: low = mid 80s, avg = 110s 5950X: low = mid 180s, avg = 220s What? Was there any recent perf patch or just Zen3? 😳😳😳
12
3
69
OMG, I’ve just been asked and have accepted it! It’s until death or TDP tear us apart, @IntelGraphics!!! 💍🤵🏻‍♂️💻❤️ #SiliconGang
9
6
73
AVX512 being what it is, AMD has done the best and only sane implementation that there is for client CPUs. Crazy times we live in and another humiliation for Intel. By the way, what the hell is Intel's messaging on the future of AVX512 on client at all? Silence? #SiliconGang
12
5
70
I love this so much!!! But it’s about to get refreshed for RDNA3! #SiliconGang teespring.com/shop/waiting-o…
8
7
67
Honestly, the M1 is unbelievable when you’re using it. Even when it’s swapping at more than 2GB (8GB RAM model), the responsiveness is equal or better than my 11700K desktop. Silent. Doesn’t really warm up. Meanwhile, the Dell Latitude 7300 with an i7 8665U is a burning jet.
11
4
63
This GPU gen is more mixed than I had initially thought. Raster: Tie. Efficiency: AMD by far. RT: NVIDIA by far. Bet on DLSS-like: NVIDIA by far. Bet on short-term VRAM usage increase: AMD by far (until Ampere refreshes). Compute: NVIDIA. Software: I think NVIDIA. Value: 🙅‍♂️
9
3
62
Will IPv6 ever be widely deployed?
21
3
58
NVIDIA has got the best GPU talent in the world — and it's plain to see when lithography isn't a variable. Period. #SiliconGang
13
5
64
Not impossible that AMD may eventually acquire Intel’s chip divisions if 2025 and 18A put Intel in even darker waters. It would be both sane and insane if it comes to that — the only company guaranteed to be genuinely interested in the continuity of x86. #SiliconGang
27
2
59
12,138
Would be fun if NVIDIA were to buy Intel. #SiliconGang
28
4
55
10,145
Hahaahahahahahhhahaaaaaahhahaahshahahahahahahahahahahahaahbshahahahsbxhwuwueiejdndhuqhahahaahanfjdieuebshahahahahabsheueheuhahahahahabsbdheuwhahahhsgetwtwtwrrqrqrqtwtwvzvxvxvxvdnfngnvngnvogoroeieueueuuwwhwhwbbwhsushsbenrjruyrhensnshhshsbdnfkgohohotlrkwnwnuwywywhsjsh. #SiliconGang
23
4
56
Would you like to hear a TCP joke, #SiliconGang?
1
10
57
It will be fun to watch M4* steamroll “high performance” x86 desktop CPUs in a wide range of benchmarks that don’t get called memes nor use “mUh SmE”. Then it will be about “nOt ReAl OS”, “expensive”, and the lovely “wAiT fOr ZeN6”. #SiliconGang
8
4
56
7,219
AMD Ryzen 7000 summary: - 60% power jump (143W to 230W) for an avg of ~45% better MT perf - AVX512-accelerated workloads and corresponding speedups (lower than Intel's) are suddenly relevant, as prophesized - Worse than 5800X3D in gaming - ~same perf, >>cost vs RPL #SiliconGang
12
6
51
Incredibly low energy presentation from AMD. It's as if AMD people themselves were bored — which is weird, because the content is wildly more interesting than RDNA2. #SiliconGang
9
3
52
Had fever again last night. Dreamt that Uncle Jim Keller was staying at my house and sleeping on my bed. I went to my Mom’s bedroom and whispered her, “Shhh, one of the world’s top CPU architects is staying here for a while and sleeping on my bed, OK?” This disease is sick.
14
1
57
Jesus Christ, we’ve just met the man, the engineer, and the legend @jimkxa thanks to @amybarton. You wouldn’t believe how cool he is — and he’s in such good shape that @Jack_Mangano just had to ask him *the* question... #SiliconGang
6
11
55
I hear whispers of McDonald’s thinking of acquiring Intel because of their chips. Whose media guys should I share this with to be in a news article? #SiliconGang
10
3
53
3,590
Time to let HDMI rot in peace. #SiliconGang
"The HDMI Forum forbids the implementation of HDMI 2.1 features on open-source drivers" Excuse me? WHAT?!
5
52
+ X86S + APX + AVX10 + FRED + RibbonFET + PowerVia = x86 and Intel back in the game, baby. ❤️‍🔥😍 #SiliconGang
9
7
54
6,099
Oh my, I knew the RTX 4000 SFF Ada would be small, but I didn't know it would be TINY! Even the A750 is a monster next to it. I won't be going for consumer GPUs again after the experience with the specs, feature set, software, form factor, and all else from this beauty. 😍❤️ #SiliconGang
7
2
53
8,082
Both Zen5% and Arrow Lake are looking so mediocre an “upgrade” outside of microbenchmarks and corner cases that it hurts. Even Apple’s yearly laughable “refreshes” are way more relevant than x86 stuff. #SiliconGang
12
5
50
10,324
OK, I’ve done everything in life. Thanks. #SiliconGang
6
6
48
Back to being a x86 shill, guys. Let's be frank, x86 is where it's at nowadays. Arm has finally hit the wall and even Intel (God forbid, Intel!!!!) does way bigger IPC increases nowadays. I re-welcome our old-new eternal silicon & ISA overlords. #SiliconGang
9
1
53
Interesting stuff. While looking into ASUS Z790 BIOS options, I noticed a setting called "CPU DLVR Bypass Mode Enable". Then, after looking for any info on it, Shamino (ASUS BIOS developer) explained [RPL-S] initially used DLVR but midway in development abandoned it. #SiliconGang
10
8
50
Apple’s in-house Mac silicon, AWS’ Graviton, and Fujitsu’s A64FX initiatives are a clear sign of the writing on the wall. Intel has done irreversible damage to x86 — that not even a hot AMD can fix — and hugely accelerated the migration to Arm by 5-10 years with the 10nm fiasco.
11
5
49
Tim Cook's Apple seems to combine some of the most meh parts from Intel/AMD. What's up with the turtle cadence? No M2 yet, seriously? After nearly no perf. increase in A14 -> A15? Just bigger SoCs with mOaR cores used by applications that can be counted with two hands?
18
6
52
OK, I’m sold into the RDNA2 hype. Might be an incredible deal. Please don’t screw the drivers, AMD. Please.
7
1
47
Replying to @mooreslawisdead
Please, spend extra 5 minutes studying — actually, even mindless reading will suffice — about the state of the UE plugin. Just read a game dev’s tweet or something and don’t publicly embarrass yourself.
48
What peak performance in t-shirts looks like! ❤️🔥 @IntelGraphics #JoinTheOdyssey
8
2
48
8GB VRAM. OK, I flop: it’s ridiculous in a 2020 high-end GPU. Nothing “high-end” should have less than 12GB right now. I am using 1.2GB VRAM just sitting idle at Windows with Steam, Plex, and common stuff in the system tray.
10
2
49
ARE YOU READY FOR MY F-I-V-E GIGAHURTZ, #SiliconGang?
11
1
49
Forrest Norrod saying Milan/Zen3 is leadership in all performance workloads, no exceptions/excuses (25:30). 🙂 #SiliconGang
5
6
47
AMD is giving emphasis to gaming in relation to Zen3/*CPUs*. Speaks volumes about how much of a winner they have in their hands this time. THAT FABLED GAMER MOMENT HAS COME. ASSEMBLE, #SiliconGang! OCTOBER, 8TH!
10
2
44
MLID is indeed so very dead. 😵‍💫😱🙃 #SiliconGang
Pat Gelsinger Becomes First Owner of Intel Arc A770 Graphics Card tpu.me/cu2k
18
2
45
What a crappy announcement of the A15. Is Apple finally hitting strong the IPC wall? Brain drain after key top architects and designers left for NUVIA and others? #SiliconGang
12
5
46
Stability, performance, sheer software compatibility & WATTS are once again the law in this house. #SiliconGang ☀️🔥🥵
14
4
48
6,078
Oh, the consumer friendly silicon company has just killed HEDT now that it hasn't had competition for a while. How the world spins round! #SiliconGang
Shed a Tear for HEDT: AMD's Official Threadripper Pro Pricing Marks the End of an Era trib.al/0sxzpdd
15
3
47
The AMD CPU + GPU performance boost thing is more damaging to Intel than it is for NVIDIA. If you’re not a big reviewer/outlet with lots of HW/time/money, and assuming Zen3 is consistently >= Skylake at games, it only makes sense to use an AMD platform for testing from now on.
6
3
43
Ei, pessoa apaixonada por open-source, DevOps e Golang! 😃💻 Que tal ser Tech Lead de um dos projetos de DX/DevOps mais incríveis nascidos no Brasil, o Ritchie CLI? 😍 Ó: ele é 100% aberto e nossa missão é revolucionar o jeito de contribuir com OSS e trabalhar em comunidade! 🥳
2
8
45
The current [un]availability of RDNA2, Zen3, and Ampere cards should serve as good indicator it’s absolutely unfeasible for Intel to leverage TSMC and/or Samsung in any significant way for their main products even in the medium-term. There’s no option but to fix their own fabs.
7
4
40
Major kudos to our @IntelGraphics friends for the Herculean effort that has finally come to fruition as a solid re-entry into discrete GPGPUs! Would love to test one of these! 😍 Thanks for this Odyssey, @RajaXg @gfxlisa @theBryceIsRt @ryanshrout @rogerdchandler @bobduffy! 🤟🏻
Intel Arc A770 and A750 Tested and Reviewed, also RTX 4090 cards in hand nitter.app/i/broadcasts/1kvJpmDOQ…
2
4
47
So, apparently, when Intel CPUs are under minimally reasonable baselines, even Zen4 SKUs already surpass them. Poor mid 5GHz Arrow Lake will be bulldozed. #SiliconGang
9
3
43
6,621
Replies to this thread make me think Tesla fanatics are more toxic than AMD ones.
Replying to @PatrickMoorhead
Here’s me driving the Tesla Model Y on the left. It, of course, doesn’t have LiDAR. On the right is the @luminartech enabled vehicle. $LAZR
10
2
39
If Zen3 is high double-digit IPC increase + slightly better clocks and it does come to AM4 with Ryzen 4K, then it won’t matter if RKL-S is Sunny or even Willow on 14nm: it’ll be a bloodshed. And if it’s still SKL, Intel will be wiped out of DIY and high-perf desktop. #SiliconGang
13
3
41
It’s over, guys. All good Zen things were Jim Keller’s. #SiliconGang
11
3
44
6,507
There is no major hidden performance in Arrow Lake, but there’s definitely major hidden consistency. Phoronix’s Linux results put ARL in an extremely consistent light, always above or — in super few cases — within margin of error vs 13/14th Gen. There should have been much better coordinated work between Microsoft and Intel for this launch. #SiliconGang
2
4
44
4,407
Wow, Gracemont is incredibly tiny next to Golden Cove. Really neat. #SiliconGang
First glance on the package of #AlderLake und #SapphireRapids? So Intel is confirming four dies for Sapphire Rapids here, but is not showing the HBM variant. Alder Lake rendering is showing the difference IP blocks used. You can spot 8+8 big and small cores.
4
2
44
Going full offline for a couple weeks. Big life changes ahead that need full focus. Wish me luck. I wish you well. 🙏🏻🌊⚓️👑
10
44
Guys, who wants to see DLSS-like upscaling/enhancement of videos in something like VLC/Plex? 😍💕 The idea was already pitched internally at NVIDIA and we need to show interest to make it happen! 👋🏻👋🏻 Please leave a comment and like here if you’d like to have it! #SiliconGang
Oh I've pitched it already. Now I just need users requesting it to motivate them :)
10
6
41
AMD has the best client AVX512 implementation, indeed. Intel needs to stop the madness and fix things with Meteor Lake. #SiliconGang
An @AMD @AMDRyzen 7 7700X vs. Core i9 11900K #AVX512 Performance Analysis -- Tons of AVX-512 on/off comparison metrics for many workloads, looking at power consumption impact too. After initially being wary, I've been loving the Zen 4 AVX-512 perf! phoronix.com/review/zen4-avx…
4
5
42
AMD is really getting into TSMC's Big Boys Club starting in H2 2020. No more French FUD about volume, "not seen as a big strategical partner by TSMC", etc. Zen3 and Milan are the real deal. #TheBestIsYetToCome #SiliconGang
Replying to @chiakokhua
@SudipChahal asked me to translate this article originally posted on Reddit. Here it is:
2
4
40
DLSS-like is the future! #SiliconGang
Sony Patents a DLSS-like Machine Learning Image Reconstruction Technology. This feature could push the PlayStation 5's performance to new heights. overclock3d.net/news/softwar…
2
1
39
To put any doubts about the funding of that paper on AMD CPU vulnerabilities to rest, the new LVI research has been funded by Intel, AMD, and ARM. #SiliconGang
3
13
40
WHO'S PUMPED UP? MAMA SHALL DELIVER THE HAMMER — PAPA WILL NEED INTERPOLATION TO SEE WHAT HIT HIM!!! 🥵😮🥺 #SiliconGang
Join us on November 3rd as we launch RDNA 3 to the world! More details to come soon! #RDNA3 #AMD
11
10
41
Have you noticed a first in that virtually no actual Zen5 scores have leaked yet (just “a source told me X points in Y”, “Jim Keller’s slides projects Zen5 Epyc at Z” etc.)? 😋 AMD has been *tight* about it. Nobody has dared to be sloppy about it. The true Bulldozer is almost here. 🔵🚜🔴 #SiliconGang
13
1
41
9,813
Zen5 is a 1T/single thread monster. Word. I might have to give Mama another chance. #SiliconGang
5
2
41
6,456
WHO’S READY FOR [FLACCID] ARROW [2%] LAKE??? #SiliconGang
11
1
39
5,352
NOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOOO. 4.9.
6
40
Also, top consumer CPUs = $800 now. Used to be <$330 during all of Intel’s Tick-Tock. Things are getting worse quick. (And for those who might say “but there was a big perf jump, proportionately higher than the $’s”, no — this industry exists due to faster *&* cheaper together.)
We live in a strange world when a $400 gaming graphics card is considered mid range. Back when I grew up, mid-range was $150-200
17
1
40
Tweeting in the middle of the bath. RIP Alder Lake, the king of enthusiast desktop CPUs that shall never be. It is amazing AMD is bringing so advanced a packaging to client chips when so many insisted it’d be too expensive even for most server CPUs. KNEEL TO MAMA. #SiliconGang
2
5
38
Well, it seems Intel Arc and/or RDNA3 will be my daily driver(s) going forward. Burning or not, I am not interested in power adapters that require 3x 8pin or in upgrading my formidable ATX 2.4 PSU. Also, yet another generation of retarded VRAM capacities. NVIDIA is losing it. 👎🏻
3
6
38
Matches not only @RedGamingTech’s sources, as claimed by the article itself, but also @BitsAndChipsEng’s (for individual integer & floating point ops) and what I’ve heard from #SiliconGang members with secret sauce. Zen3 = better ~10-12% INT, ~50% FP, ~17% mixed.
#AMD Zen 3 Core Rumored To Bring Massive 17% IPC Gains In Next-Gen #Ryzen 4000 & #EPYC Milan CPUs wccftech.com/amd-zen-3-ryzen…
10
3
36
19% better IPC, reaching ~>TGL levels. Only a year after Zen2 and a 15% jump. What AMD did in ~14 months, Intel couldn’t do even in ~36 — and that’s *disregarding* 10nm woes, because SKL -> TGL is a 3 uArchs jump (CNL, ICL, TGL), with the last being a... ~3% regression over ICL.
7
2
39
Person really close to me has just got tested positive for COVID-19. Think I had some light fever at some points in the last ~2 days, and today I have some headache and light body discomfort (coughing, too, but I’m very allergic, so). Pretty confident I’m positive as well...
49
2
38
Here’s a great article about @AMD’s @LisaSu’s beautiful professional trajectory and amazing leadership skills, including an “Apollo 13-like” mission to save the first Zeppelin that came back from the foundries from an unexpected design flaw. Thanks, Lisa! fortune.com/2017/06/28/amd-a…
4
7
36
KNEEL. 🙏🏻 #AMDomination #SiliconGang
$AMD announces strong second quarter results and increased 2020 revenue guidance. Learn more: bit.ly/2Eq3CsZ
3
1
36
Intel HEDT is dead until what, 2023?
10
1
36
Buddy from Brazilian retail channel said there weren’t even 100 Ampere GPUs in the hands of BR distributors.
5
4
37
How is AMD so unable to make USB work, after AM4 and AM5? 🤢 #SiliconGang
22
4
39
5,925
If Intel had truly decent 10nm numbers to show, they’d talk absolute numbers instead of the “X percent better volume than what we expected” meme that has been going on for many quarters already.
4
1
37
TGL regressed in IPC compared to ICL, haha. It’s 97% of ICL. #SiliconGang
3
2
36
So, after much reflection and not wanting to be unfairly negative to it and AMD, I have compiled a list of Zen5’s biggest strengths as a client uarch and desktop product, #SiliconGang. A thread. 🧵👇🏻
6
3
39
5,699
Replying to @JubileuGomes6
Actually, those tests are from the original review and, as per their methodology, only change the PL1/PL2 values. The “Intel baseline” is not limited to that — it lowers Icc Max, forces CEP to be enabled, increases the average voltage, and other things. At the old defaults, 13900K and so on easily reach ~350W several times during a CB MT run, for instance.
6
1
37
2,458
Never install ASUS Armory Crate. It really builds an armory crate of junk all over your Windows install and you’ll never fully remove it.
8
1
33
Dreams come true! US Embassy Brazil has approved my visa and will expedite it this Thursday, so that I’ll be able to join the awesome silicon heroes and heroines from @IntelGraphics — with quite a few lovely silicon gangsters — at their meetup and GDC next week! Blessed! 😃🥳🙏🏻❤️
7
2
34
Jim Keller Facts: Whereas mortal people were using that gun thing to shoot t-shirts across the room at @IntelGraphics’ Odyssey event, @jimkxa just picked one up with his right muscle and threw it 2x farther than the gun shots, giving a satisfied smile at the end. #SiliconGang
4
2
36
Absolutely loved @AMD’s presentation style. Felt like a movie trailer, no nonsense, just the bits we wanted from the old 1:30 hour long unbearable events. Congrats, @LisaSu!
4
1
36
AMD needs to up their game and have at least the top desktop Ryzen SKU be in the leading edge node of the moment, otherwise it will get severely behind Intel in all upcoming generations. This “using previous gen stuff because cheaper” won’t do anymore going forward at the high end. Desktop Zen4+5 will make them be at 5nm for >~4 years before Zen6 comes out, while most others moved to 3nm this year (not Qualcomm, but X Elite is also a kind of dud because of it). #SiliconGang
15
2
38
7,592
Zen5% has lower ST performance and usually even MT than Intel’s recycled RPL refresh, lol. What a fail. Terrible uarch. #SiliconGang
9
4
40
7,644
AMD's Ryzen 7 5800X3D is, simply put, a bad final product that shows V-Cache has no widespread future in client SoCs except maybe in iGPU configs — and even then, with highly questionable cost/benefit that will likely kill it. #SiliconGang techpowerup.com/review/amd-r…
25
4
38