*sigh*

Reading and typing on my 2013 dual core i5/8GB.

You see, MOST people really DON'T need/use advanced instruction sets or CPU/GPU's in the daily lives. (MAYBE at 'work', but that is another story...)

Quote from: Tribbls on July 21, 2024, 15:29:23I'm still using an HP Probook(4540s) laptop that's got a socket packaged Intel Ivy Bridge 3632QM "Mobile" Processor. And so that Laptop's BGA soldered to the MB Radeon HD 7650M(Terascale Rebrand) dGPU has given up the Ghost and I could get a new MB and swap the CPU over to that as the old parts for that Business Grade laptop are still in good supply owing to the fact that corporations had fleets of these laptops to manage. But the Laptop sill works fine with the Ivy Bridge Intel HD 4000 integrated graphics and so I've never bothered getting a new MB to fix that dead discrete GPU. I can still go over to the Local Micro Center and get a replacement battery for that as the laptop has the removable battery that can be swapped/replaced when needed for longer battery life with a spare battery or easy replacement with a new battery as I have done when the old battery expired!

I have an old HP Elitebook 840 G1 which has HD 8750M, best laptop I have ever owned. I am typing on it right now even and it works excellent even to this day. Too bad HP business laptops have went from some of the best to junk. I am still amazed also how easily serviceable this laptop is

QuoteBut that's only because of AMD's non existent ROCm/HIP support for Polaris Graphics, or even ROCm/HPI support for Vega iGPUs! And so no Blender 3D iGPU/dGPU accelerated Cycles rendering on Linux Mint and Polaris graphics was dropped from the ROCm/HIP support matrix years ago and Vega Graphics is on borrowed time for that ROCm/HIP support matrix soon as well.

AMD doesn't officially support it, but you can get ROCm/HIP running. I did on my old Polaris card in Mint

Quote from: lol on July 21, 2024, 16:37:17We trust you bro, except my PHONE connected to a screen, keyboard and mouse with a cable and full Arch Linux + xfce4 does 1900/5500 geekbench and the other one 1300/3400 geekbench.

When I see beginning of posts like this it's clearly yet another retrograde who only does very basic office work on his laptop and could actually use a tablet instead.

I'm going to disagree with you. My computer is just 1 year older(but it is a U processor, I have better single core performance but worse multicore, 866/1617 vs his 500/1677) and I do programming on it all the time

The reality is for most usage, it isn't that big of a deal. Only time I send stuff to my faster desktop pc is when I need to do compiling. But for interpreted stuff, it works good enough that I have little stress.

Synthetic benchmarks are just that, synthetic. It helps when you aren't on windows with billions of things running in the background. It also helps that browsers these days suspend background tab processes

The real thing that hampers old stuff is instruction sets. Like if you have a processor before AES instruction sets or SSE/AVX. Or h264 hardware acceleration. These are the big stuff, the every day stuff you use in every day life that when offloaded to the processor can be a pain. With these instruction sets even slower cpus aren't effected as much for most every day use.

It's like getting an SSD, the slowest SSD is a huge upgrade over an HDD. But the fastest NVME is an improvement but not as much compared to the jump of the above. That is pretty much how computing is, it is why budget android phones used to be crap but these days are good enough for most people. It is also why intel intentionally kept many instruction sets away from their lower celeron/pentium/i3 lines. To make the performance bad enough that it is usable but you see an upgrade.

Interestingly enough, many people aren't even using the full potential of their hardware. One of the reasons why chromebooks are able to get decent performance out of low end hardware is because they take advantage of this fact. It is based on gentoo where the software is compiled and optimized for the hardware making full advantage of the hardware

Quote from: Tribbls on July 21, 2024, 18:09:04And the 3632QM does fine with Light Blender 3D workloads and plenty of other graphics applications

Riiiiight

Quote from: Tribbls on July 21, 2024, 18:09:04Old Intel processor(3832QM) does just as good as My newer AMD Ryzen 5 3555H laptop

Correct way to say it is "they are both around the same level of bad old slow hardware".

Quote from: Tribbls on July 21, 2024, 18:09:04No Linux Tablets that are running full Linux OS distros

Google "termux proot linux".

Quote from: Tribbls on July 21, 2024, 18:09:04because not all the processor makers have that properly supported, cough cough AMD

Correct way to say it is "Blender doesn't support it on every hardware". AMD has nothing to do with lazy Blender devs.

"another retrograde" NOT! And the 3632QM does fine with Light Blender 3D workloads and plenty of other graphics applications.  But really that Old Intel processor(3832QM) does just as good as My newer AMD Ryzen 5 3555H laptop and the Vega 8 iGPU and that ASUS Tuf laptop's Radeon RX 560X dGPU! But that's only because of AMD's non existent ROCm/HIP support for Polaris Graphics, or even ROCm/HPI support for Vega iGPUs! And so no Blender 3D iGPU/dGPU accelerated Cycles rendering on Linux Mint and Polaris graphics was dropped from the ROCm/HIP support matrix years ago and Vega Graphics is on borrowed time for that ROCm/HIP support matrix soon as well.

And so because of some lack of iGPU/dGPU compute API support on Linux from AMD, and even Intel way back at Ivy Bridge time, I'm forced to Blender Cycles render on the CPU cores instead of the iGPU/dGPU on the laptops. And Ivy Bridge was the first generation of Intel Processors that had OpenCL support for the iGPU but at that time neither Intel nor AMD where making much effort to get their processors with iGPUs working properly with Blender 3D, even back when Blender 3D still utilized OpenCL as the iGPU/dGPU compute API.

But the Blender Foundation has since dropped OpenCL(Since Blender 3.0 was released)  as the iGPU/dGPU compute API in favor of Nvidia's CUDA or Apple's Metal.  But Intel's OneAPI/level-0 is better supported on iGPUs/dGPUs on Linux for Intel's Xe/Xe-LP(ARC) graphics whereas AMD's ROCn/HIP is not well supported on Linux. And so Modern Intel Graphics works fine for Blender 3D oGPU/dGPU accelerated Cycles rendering but not AMD's  Polaris dGPUs or Vega iGPUs, currently, if they ever will be supported. And ROCm/HIP(AMD) And OneAPI/Level-0(Intel) are what's taking the CUDA(PTX) that Blender 3D is generating and converting that PTX(Portable Intermediate Language Representation) to a form that can be executed on AMD's and Intel's respective iGPU/dGPU hardware. 


As far as Tablets are concerned No MacOS support on Apple Tablets and No Linux Tablets that are running full Linux OS distros means that I'll never purchase any tablet. And I'd love to get an Apple Tablet that's using an M series processor because Apple makes the effort to make sure that their MacOS hardware has the proper iGPU compute API(Metal for Graphics and GPU compute) support to work with Blender 3D's iGPU/dGPU accelerated Cycles rendering so folks do not have to be forced to use that fallback CPU cores accelerated Cycles rendering that's just too stressful on laptop cooling solutions to be used regularly.

Sure the Tech press are the only ones really using Blender 3D's CPU cores accelerated Cycles rendering for stress testing CPU cores but even iGPUs can make quicker work of rendering than any CPU cores! But the Tech Press is Totally Ignoring any Blender 3D iGPU Accelerated Cycles testing because not all the processor makers have that properly supported, cough cough AMD. And so we get no Intel Xe/ARC Blender 3D iGPU related Cycles rendering tests and Nvidia's mostly dGPUs that have always had Blender 3D's GPU accelerated Cycles rendering supported first and for the longest time. And AMD's needing some shaming there to get them to support their iGPU/dGPU hardware for Compute APIs on Linux.

Quote from: Tribbls on July 21, 2024, 15:29:23Intel Ivy Bridge 3632QM

500/1600 geekbench

Quote from: Tribbls on July 21, 2024, 15:29:23Laptop sill works fine

We trust you bro, except my PHONE connected to a screen, keyboard and mouse with a cable and full Arch Linux + xfce4 does 1900/5500 geekbench and the other one 1300/3400 geekbench.

When I see beginning of posts like this it's clearly yet another retrograde who only does very basic office work on his laptop and could actually use a tablet instead.

I'm still using an HP Probook(4540s) laptop that's got a socket packaged Intel Ivy Bridge 3632QM "Mobile" Processor. And so that Laptop's BGA soldered to the MB Radeon HD 7650M(Terascale Rebrand) dGPU has given up the Ghost and I could get a new MB and swap the CPU over to that as the old parts for that Business Grade laptop are still in good supply owing to the fact that corporations had fleets of these laptops to manage. But the Laptop sill works fine with the Ivy Bridge Intel HD 4000 integrated graphics and so I've never bothered getting a new MB to fix that dead discrete GPU. I can still go over to the Local Micro Center and get a replacement battery for that as the laptop has the removable battery that can be swapped/replaced when needed for longer battery life with a spare battery or easy replacement with a new battery as I have done when the old battery expired!

And Intel Socket Packages some of it's mobile processors every year in the form of those T/Other series processors that are 35W and used on those Dell/HP/Lenovo Thin Client Mini Desktop PCs for businesses to utilize. And even Intel's Meteor Lake processors have some socket packaged processor options for OEMs/Embedded Market usage. And all of AMD's "Desktop" APUs are Mobile variants packaged for Socket Based MBs instead of BGA/Soldered packaging. And so All of AMD's Ryzen "Desktop" APUs come from Mobile oriented Die samples and that lack the latest PCIe(5.0) generational support.

Intel's Mobile processor platform I/O lanes support is more like Intel's desktop Platform for PCIe lanes generational support(PCIe 5.0) whereas AMD's mobile platform I/O lane generational support is always trailing AMD's Desktop Platform I/O Lane support! And with USB4-V2 and TB5 incoming PCIe 5.0 lane support will become more necessary if one wants sufficient USB4-V2 or TB5 Ports on laptops! 

But At least on the Framework Laptop the Old MB module with its BGA soldered processor can be sold on the second hand market or repurposed into a mini Desktop PC with that CoolerMaster external case or some 3D printed case and avoid becoming landfill!
 

Quote from: ghm on July 18, 2024, 17:28:37Except cpu sockets get outdated faster than you will buy your next laptop.

You are kinda right with that. In the past that was definitely the case. But now with the computer industry more focusing on APUs, any CPU upgrade would also be an iGPU upgrade. And in that case, socketed CPUs (even in laptops) would be great.

Except cpu sockets get outdated faster than you will buy your next laptop.
That upgradeable 2017 Sky X4C laptop with 1151 couldn't offer anything in 3 years, Core i9-9900KS - about the power of 2020 M1 - but hotter, way bulkier and crappier overall?

Sock it!

Yeah, socketed CPUs would easily work on laptops. But suddenly they said "blabla... too hot, no space for cooling", which was just an excuse to remove modularity from laptops. If you look at the CPU lineup each generation, they always have like 20 CPUs in different specs and power configurations. Some of them would always work for laptops.

Also, tiny PCs like the AsRock Jupiter are quite thin and also have socketed CPUs. That's not much different to a laptop.

And if the chip is really too hot for the case, then just don't put it in. And before someone says: "but people could put in high end chips, which may produce excessive heat and destroy something, and still demand repairs and warranty" - that could easily be prevented, as the allowed CPU types can be limited in the BIOS.

Finally, they could also standardize motherboard sizes for laptops, e.g. like 3-5 standardized sizes, depending on the laptop size. Would be reasonable, right? But the industry has no interest in that.

They should standarized CPU and GPU socket for mobile like sodimm or NVME

The "problem" lies with the technology involved.

Granted a dGPU is a PCIe device and much like the desktop counter parts, give it a PCIe socket/slot and a way to pipe enough power into it and you could have a swapable module.

CPU's however on the other hand are to much of a mixed bag. Not only might the number of pins and pinouts themselves different between models/generations the number of (and voltages for) the supply rails can be wildly different as well.

Try to "boil that down" to some sort of exchangeable/replaceable "module" that might be good across more than 1-1.5 generations and you'll see the point.

Long long long gone are the days where the CPU could merely be swapped out with a 25/33/66/99Mhz part with the same pinout. :)