Digital_Jedi
@digital-jedi.bsky.social
Photolithography researcher at IFS and tech enthusiast.
God imagine the horror of pulling up your cooler for a repaste and watching it rip a memory chip off because the pads were just that sticky.
December 1, 2025 at 9:12 PM
God imagine the horror of pulling up your cooler for a repaste and watching it rip a memory chip off because the pads were just that sticky.
They're always so beautiful in the wafer. I'd love to see the quartet of wafers that makes a PTL chip all together. CPU, GPU, I/O, and Base Tile all on different nodes, and 2 options for the GPU.
October 15, 2025 at 9:27 PM
They're always so beautiful in the wafer. I'd love to see the quartet of wafers that makes a PTL chip all together. CPU, GPU, I/O, and Base Tile all on different nodes, and 2 options for the GPU.
56% lower power and 2x performance is insanely impressive. Over 4x efficiency improvement.
September 22, 2025 at 7:39 PM
56% lower power and 2x performance is insanely impressive. Over 4x efficiency improvement.
100%. Intel needs to show a goal and path forward. 18A and 14A are nice, but we need to see a commitment to taking the good bits of the current lineup and carrying them forward.
Things like Lunar Lake and ARL-HX are winning formulas, and ARC has potential. PTL, NVL, & Xe3 need to stick the landing.
Things like Lunar Lake and ARL-HX are winning formulas, and ARC has potential. PTL, NVL, & Xe3 need to stick the landing.
July 25, 2025 at 5:22 PM
100%. Intel needs to show a goal and path forward. 18A and 14A are nice, but we need to see a commitment to taking the good bits of the current lineup and carrying them forward.
Things like Lunar Lake and ARL-HX are winning formulas, and ARC has potential. PTL, NVL, & Xe3 need to stick the landing.
Things like Lunar Lake and ARL-HX are winning formulas, and ARC has potential. PTL, NVL, & Xe3 need to stick the landing.
AMD survived the Bulldozer era. Time for Intel to take their lumps and either come back stronger or die trying.
The industry gets a lot darker without them in it, so I think it's in all our best interests to hope they pull through.
The industry gets a lot darker without them in it, so I think it's in all our best interests to hope they pull through.
July 25, 2025 at 5:10 PM
AMD survived the Bulldozer era. Time for Intel to take their lumps and either come back stronger or die trying.
The industry gets a lot darker without them in it, so I think it's in all our best interests to hope they pull through.
The industry gets a lot darker without them in it, so I think it's in all our best interests to hope they pull through.
I have it set to send all notifications for your channel, but still didn't get one for the last video.
July 21, 2025 at 12:53 PM
I have it set to send all notifications for your channel, but still didn't get one for the last video.
Yeah for sure. Trade that area for doubling the iGPU into something more usable.
July 16, 2025 at 3:55 PM
Yeah for sure. Trade that area for doubling the iGPU into something more usable.
They really need to bring back the Ryzen 3 name for things like this.
July 16, 2025 at 3:54 PM
They really need to bring back the Ryzen 3 name for things like this.
It's such a chicken and egg problem too. No laptops support it because nobody makes any modules, but they don't make modules because there's no products to put them in. We need some brand to just decide to go all-in on it and work with some manufacturer to scale up.
June 30, 2025 at 11:48 AM
It's such a chicken and egg problem too. No laptops support it because nobody makes any modules, but they don't make modules because there's no products to put them in. We need some brand to just decide to go all-in on it and work with some manufacturer to scale up.
I really hope we see some CAMM2 ram in thinkpads next-gen. Faster transfer rates to feed the big iGPUs of modern SoCs and still replaceable by the end user.
June 29, 2025 at 9:51 PM
I really hope we see some CAMM2 ram in thinkpads next-gen. Faster transfer rates to feed the big iGPUs of modern SoCs and still replaceable by the end user.
I like that you can make out what looks like 44 cores on each one here.
Given the flagship 6980P has 128 cores on 3 of these, that would mean 4 total are disabled. 2 tiles with 1 each and 1 with 2 most likely, though it could be one full and 2 with 2 each disabled.
Given the flagship 6980P has 128 cores on 3 of these, that would mean 4 total are disabled. 2 tiles with 1 each and 1 with 2 most likely, though it could be one full and 2 with 2 each disabled.
June 18, 2025 at 2:18 PM
I like that you can make out what looks like 44 cores on each one here.
Given the flagship 6980P has 128 cores on 3 of these, that would mean 4 total are disabled. 2 tiles with 1 each and 1 with 2 most likely, though it could be one full and 2 with 2 each disabled.
Given the flagship 6980P has 128 cores on 3 of these, that would mean 4 total are disabled. 2 tiles with 1 each and 1 with 2 most likely, though it could be one full and 2 with 2 each disabled.
If you did it really, really carefully, it might become a new way to do that though.
May 2, 2025 at 4:22 PM
If you did it really, really carefully, it might become a new way to do that though.
I'll be very curious to hear your thoughts on the new P/E-core placement scheme. Splitting up the hottest cores has seemed to help with hotspot temps comparing my 13900K and 285K, but I wonder how much of that is also just ARL being much more efficient than RPL and kicking out less heat.
April 21, 2025 at 4:43 PM
I'll be very curious to hear your thoughts on the new P/E-core placement scheme. Splitting up the hottest cores has seemed to help with hotspot temps comparing my 13900K and 285K, but I wonder how much of that is also just ARL being much more efficient than RPL and kicking out less heat.
I like how it just ends. Looks like they gut a graphics card in half.
April 16, 2025 at 1:32 PM
I like how it just ends. Looks like they gut a graphics card in half.
Kinda wild to me that we just casually talk about Tb/s interconnects outside of a chip's internals now. That's going to be some crazy networking design to scale out to 9k chips as well.
April 14, 2025 at 4:04 PM
Kinda wild to me that we just casually talk about Tb/s interconnects outside of a chip's internals now. That's going to be some crazy networking design to scale out to 9k chips as well.
Out of curiosity, why shoot 4 bullets at once, but not have a multiple of 4 rounds for that weapon? I feel like that one extra bullet is just kinda sitting there. I feel like either 120 or 128 would make sense.
April 14, 2025 at 4:02 PM
Out of curiosity, why shoot 4 bullets at once, but not have a multiple of 4 rounds for that weapon? I feel like that one extra bullet is just kinda sitting there. I feel like either 120 or 128 would make sense.
On one hand, SW bloat is job security for me over in HW research & development land, and unused RAM is wasted potential.
On the other hand, why is Teams of all things eating 1.2GB of RAM? Like why? What could it possibly want with that much?
On the other hand, why is Teams of all things eating 1.2GB of RAM? Like why? What could it possibly want with that much?
April 9, 2025 at 1:47 AM
On one hand, SW bloat is job security for me over in HW research & development land, and unused RAM is wasted potential.
On the other hand, why is Teams of all things eating 1.2GB of RAM? Like why? What could it possibly want with that much?
On the other hand, why is Teams of all things eating 1.2GB of RAM? Like why? What could it possibly want with that much?
This wasn't supposed to be about modern SW bloat, but I do have some opinions there too. I don't think it's entirely reasonable to expect to be having a modern computing experience on <1GB RAM, but I would also love if everybody could chill with the electron repacks eating a gig each.
April 9, 2025 at 1:47 AM
This wasn't supposed to be about modern SW bloat, but I do have some opinions there too. I don't think it's entirely reasonable to expect to be having a modern computing experience on <1GB RAM, but I would also love if everybody could chill with the electron repacks eating a gig each.
This whole standard deserves to be paved over by whatever comes after. Don't built on it. Don't try to salvage it. Just start over with whatever the 4.0 connector is. At this point just give up and use an XT120 or whatever.
April 7, 2025 at 4:26 AM
This whole standard deserves to be paved over by whatever comes after. Don't built on it. Don't try to salvage it. Just start over with whatever the 4.0 connector is. At this point just give up and use an XT120 or whatever.
I guess what I'm getting at here is that I wish more people would try to truly USE the absurdly powerful machines they have access to for more things, even if those things are trivial. Go write some shitty code that does something simple. It's the first step to making a new tool for yourself.
April 7, 2025 at 4:20 AM
I guess what I'm getting at here is that I wish more people would try to truly USE the absurdly powerful machines they have access to for more things, even if those things are trivial. Go write some shitty code that does something simple. It's the first step to making a new tool for yourself.
The year I was born, and I'm going to date myself here, these sorts of calculations were impossible for the average person or would take so long you'd never do it. Now you can just write some python and know the hundred millionth number in a sequence in probably a few minutes at most.
April 7, 2025 at 4:20 AM
The year I was born, and I'm going to date myself here, these sorts of calculations were impossible for the average person or would take so long you'd never do it. Now you can just write some python and know the hundred millionth number in a sequence in probably a few minutes at most.
This video comes to mind showing just how fast a computer is. The vast majority of the time spent running that program comes from putting the results on the screen and save it to a file.
www.youtube.com/watch?v=cmsh...
www.youtube.com/watch?v=cmsh...
Computers are REALLY fast
YouTube video by Softwave
www.youtube.com
April 7, 2025 at 4:20 AM
This video comes to mind showing just how fast a computer is. The vast majority of the time spent running that program comes from putting the results on the screen and save it to a file.
www.youtube.com/watch?v=cmsh...
www.youtube.com/watch?v=cmsh...
That cache design looks like it drove some poor layout engineers insane. With N44 being allegedly so much smaller, I wonder how much it will look like half of N48.
April 2, 2025 at 6:14 PM
That cache design looks like it drove some poor layout engineers insane. With N44 being allegedly so much smaller, I wonder how much it will look like half of N48.
Very excited to see a new Doom gaming coming and the shield mechanics look really interesting. I had fun with both Doom 2016 and Eternal, but this looks like a cool change too.
April 2, 2025 at 6:12 PM
Very excited to see a new Doom gaming coming and the shield mechanics look really interesting. I had fun with both Doom 2016 and Eternal, but this looks like a cool change too.