Guido Vandecasteele (Fully Buffered) (@FullyBuffered) header
Guido Vandecasteele (Fully Buffered) (@FullyBuffered) profile picture

Enthusiast of obscure PC hardware, CrossFit, Consultant at CoolProfs. Buying too much stuff and then forget to sell it. Check out my Youtube page!

The Netherlands
Joined May 2011
Photo by Guido Vandecasteele (Fully Buffered) (@FullyBuffered)Photo by Guido Vandecasteele (Fully Buffered) (@FullyBuffered)Photo by Guido Vandecasteele (Fully Buffered) (@FullyBuffered)Photo by Guido Vandecasteele (Fully Buffered) (@FullyBuffered)Photo by Guido Vandecasteele (Fully Buffered) (@FullyBuffered)Photo by Guido Vandecasteele (Fully Buffered) (@FullyBuffered)Photo by Guido Vandecasteele (Fully Buffered) (@FullyBuffered)Photo by Guido Vandecasteele (Fully Buffered) (@FullyBuffered)Photo by Guido Vandecasteele (Fully Buffered) (@FullyBuffered)Photo by Guido Vandecasteele (Fully Buffered) (@FullyBuffered)
When I was into overclocking, in order for it to run fast, we had to run it cold. Liquid Nitrogen cold. But going beyond to Liquid Helium temperatures, a superconducting chip could potentially run at near zero energy. That's what @SnowcapCompute is doing youtube.com/watch?v=urjzD4Vmโ€ฆ
7
6
1
40
4,222
I should have posted months ago given the hype, but here's my 65min deepdive into @substrate's claim to re-energize USA litho with an x-ray tool, supplanting @ASMLcompany as the kingmaker. It'll cost a trillion dollars, but the economics need to work. youtu.be/z_r5Bsn_DH4
6
13
5
73
8,671
Me and @ASMLcompany CEO Christophe Fouquet on stage today. Goals ๐Ÿค˜ Next stop, HQ Tour.
7
5
1
87
5,781
Can't believe it's been two years since I visited @intel's High-NA machine in Oregon. FWIW, I'm interviewing @ASML CEO Christophe Fouquet later this week on stage at @hellotmrc in Amsterdam. Should be fun
10
12
180
8,443
Were they able to wear them all in the correct location, or were a few in the wrong location because how do you wear that many watches? Also, it's only one journo over one night. Statistically insignificant. This isn't my issue with sleep monitoring. My issue is that I can have a great night of sleep, only for my monitor to tell me it 'scored 40/100' and that makes me feel like crap for the day. Or I have a bad night sleep and it scores me '82/100' and I don't trust it. Sometimes you can have too much data. In 2026, all I do is sleep. No monitoring.
A journalist wore 4 wearables simultaneously and compared them to an in-lab sleep study. The winner? Apple Watch matched total sleep time to the minute and had the closest agreement with the lab sleep-stage results. They all overestimated deep sleep.
6
2
45
6,347
Imagine if we still had Optane, what would happen? $intc and $mu to $5T
26
13
3
292
33,092
youtu.be/JdB722MK380 It looks like NVIDIA is launching the N1X tomorrow. If you've been following my tweets, it seems like a really confusing product. It's late, the power is questionable, and how much is this thing going to cost? It's 2:30am, Jensen might prove me wrong at the keynote tomorrow, but I wanted to share my thoughts.
17
16
3
141
14,095
Guido Vandecasteele (Fully Buffered) retweeted
So so sad.A lovely brilliant funny giant of a bloke. Oh Man....RIP Chris .. Boro legend forever. Love to family and friends.
414
2,186
71
37,727
966,051
Guido Vandecasteele (Fully Buffered) retweeted
The decade-old @AMD FX-9590 can run Battlefield 6 at 30-40FPS with a RX 5700, maxing CPU at 98%. Intel i7-2600K can't launch it due to Secure Boot issues. โš™๏ธ๐ŸŽฎ tweaktown.com/news/108750/amโ€ฆ #AMD #Battlefield6 #Gaming
1
1
1
482
Guido Vandecasteele (Fully Buffered) retweeted
In a new video shared on YouTube, user Fully Buffered attempted to run Battlefield 6 on two old CPUs. Despite lacking TPM 2.0 support, the 12 year old AMD FX-9590 CPU is capable of running the game surprisingly well. Can still reach upto 40s FPS range at 1080p resolution โžก๏ธ wccf.tech/1j2hv
2
8
2
39
5,548
Guido Vandecasteele (Fully Buffered) retweeted
New video! Battlefield 6 on AMD FX...it's possible I've been out to test whether BF6's minimum requirements are actually in place and made some interesting discoveries. youtu.be/bJf90cg6Olg
1
1
484
Just a note, fellow semiconductor educators. If everything is a new breakthrough, a new paradigm shift, or ushering in a new era, then your bar for true genuine innovation is too low. Your coverage should be based on the little step in context, be it academia or startup, not the little step extrapolated to a trillion dollar business that'll destroy NVIDIA or SK Hynix seemingly overnight. You don't do any favors with big whatifs. Remain practical. Identify the immediate path, not the vision of a post-silicon world overnight. Enabling an 84x84 four layer CNN today is not the same as being the thing OpenAI uses to build GPT6 or GPT7. It's the service to your audience to cut through the marketing extrapolation. I work with companies to help them tell the story of building the foundation, not how it'll look on top of a theoretical throne. Don't drink the kool-aid. Sincerely, an educator fed up with sensationalism.
15
27
4
290
23,564
Almost missed it, but @TechTechPotato is now five years old! Many thanks to the 100k+ of you that have subscribed. ๐Ÿ’–๐Ÿ™ Plenty of content brewing. I'm actually in the studio this week doing some filming!
29
10
335
17,230
Why is this a terrible headline? People are interpreting this as Deepseek's only breakthrough is that they've done it without NVIDIA's proprietary language. That's Incorrect. CUDA is NVIDIA Proprietary. PTX is NVIDIA Proprietary, even more so. PTX is already everywhere in lots of optimized deployed models. The headline avoids all this nuance, relying on people to read the article. If your headline leaves nuance on the table and misdirects, it's CLICKBAIT. Or weaponised incompetence. But you decide. It's confusing wider media and investors, and I'm having to correct them. --- FWIW, Deepseek tied several elements of research together. Mostly already in the literature. PTX is impressive, but it's run of the mill for models at the leading edge or in bulk deployment. What had me more interested was the re-writing of the communications stack and networking. That's something which isn't easy, but also might not be widely applicable.
This quoted post is unavailable.
6
28
5
287
20,564
Readers forced out by frustrated anonymous-sourced headlines
1
3
40
6,122
Guido Vandecasteele (Fully Buffered) retweeted
S3 didn't disappear after the Savage series! Quite the contrary, they would continue on under VIA with the Chrome series, starting with DeltaChrome. Then came GammaChrome, and...........no one wanted to review it. Until now! youtube.com/watch?v=_56eNWiQโ€ฆ
2
10
1
41
1,688
Guido Vandecasteele (Fully Buffered) retweeted
New video! - Thanks To This GPU, NVIDIA Became World's Most Valuable Company A retrospective of the 8800 GTX and how GPUs turned from gaming into compute devices with CUDA... youtu.be/Wjxgy8nDSso
4
2
19
1,081
Some people practice rubber duck coding to keep focus and help with problems. But it doesn't have to be a rubber duck, it can be anything. In my case, it's the @AMD TWKR Black Edition.
5
3
37
6,699
Guido Vandecasteele (Fully Buffered) retweeted
This is the Corvette that Windows zip folders bought. Back in '93 or so, I was working at Microsoft on COM, and at home for fun I started writing a shell extension to browse zip folders in the new Win95 user interface, making them appear as if they were just folders. That grew into a shareware product call VisualZIP. Then one day just before heading into work, I got a call from a lady at Microsoft. She wanted to know if I was the Dave Plummer who wrote VisualZIP, and long story short, they wanted to buy it, and could I come in to talk about it at some point? I said "Sure, what's your office number, I'll stop by!" and it kinda freaked her out. She said "No, no, we'd have to coordinate with travel and legal..." but I was confused as to why I'd need to book travel to talk to someone where I already worked! And THEN I figured it out. She didn't know that I already worked for Microsoft, and I didn't know that she didn't know. So that was a bit awkward. But we worked it out. In the end, my choices were pretty limited - either quit my day job and compete with Microsoft, or do what I wound up doing instead: cheerfully accepting their first, best, and only offer. So that's what I did! I accpepted their offer, paid the taxes on it, and bought a lightly used red 1994 Corvette LT1. There wasn't much left over. So next time you open a zip file on Windows, think of my car :-) If you've ever wondered why the zip support is so slow today, there are two main reasons: first, being 25+ year old code, it's single threaded. It doesn't matter how many CPU cores you have, it only uses one. Second, because of the way the shell used to work, you couldn't just hand it the contents of a file, you had to give it a local file path as the source. So the code first extracts the file to a temporary location, hands that location to the shell, and the shell copies the file. In other words, there's an extra temp copy operation involved in every operation. That *could* be fixed, but I imagine the sense is that anyone who's hardcore about their zip performance or feature set will likely be using 7-Zip or WinRAR, etc. I don't see them improving it any time soon, but they have added more compression formats like RAR to the mix, so you never know. If you enjoy these random nostalgic looks back, follow me for more! And heck, if you made it this far, you might as well share it too :-)
124
786
71
7,139
668,589