>traps your heat
has there been anything more retarded than this in CPU design?

>>105819685 (OP)
>xhe not heatmaxxing
ngmi

the heat of the processor cooks the code

>>105819685 (OP)
Unused temps are wasted temps

>>105819685 (OP)
How hot do you think a single core P3 god anon?

>>105822067
*got

>>105819685 (OP)
>>105819685 (OP)
have you ever had one in your hands? the other side had a heatsink and sometimes fan (Celeron or 1st Gen PII Deschutes). The heatsink was in direct contact with the CPU. it didn't even need anything else. look:

https://www.engineering-sample.com/gallery/Intel/090%20Pentium%20II/Pentium%20II%20350%20ES/

>>105819685 (OP)
>>105822139
As usual, OP is an underage faggot larping as someone who actually experienced old hardware

>>105819685 (OP)
today's cpus at idle with active cooling have hotter hotspots than that thing did in full load

>>105819685 (OP)
what heat?

I wish we returned to detachable cpu boards

>>105822139
>see? It only needs a heatsink twice the size of the equivalent K6-2/K6-3 for the same TDP

>>105824454
As much as I hated the cooling solutions of the slotted Intel CPUs, the other option of "if you slip just slightly with a screw driver under pressure pointed at your mainboard, you'll brick it" variant of K6 and Socket 370 using springy metal clamps to hold the heatsink wasn't much of an improvement. Broken die corners was also a thing...

>>105825132
What happened to socket CPUs, why aren't we seeing a return of them? (Then the PIV Prescott housefire consumed so much energy that made intel consider a whole new mobo form factor just to cool the thing down, but cooler heads prevailed and they made a dual core CPU). But now?
> Small footprint boxen are abound
> The pin count of the processors make it harder and harder to design a good socket design.
> The vast majority of a CPU's pins are VCC and GND. The new slot should have the VRM next to the cpu as it draws external 12V from a 6/8 pin gpu plug
> SMP is a thing now. In the place of a CPU socket, we can have 2 slots.
> Cooling tech improved. A lot. We can aircool 3-400W GPUs in a card format. Then aircooling a 100-150W Prescott was a challenge in a socket.
This way, the slot can focus on the memory tx/rx and pcie lanes or could have integrated memory modules to the cpu using hbm2 or good types of memory. We would upgrade cpu/ram like we upgrade a gpu now.
If ram was on board, the slot would only need pcie lanes, maybe a few lanes of displayport, and vcc/gnd in order not to get cpu gnd and signal gnd too far from the rest of the mobo. The mobo's responsibility would be to route the pcie lanes, and convert them (ie. to ethernet and sata).

>>105827211
Have you seen how many pins modern CPUs have? The socket would be a meter long.

>>105827211
Now that's something I could get behind, this would actually mean a benefit with having slotted CPUs. Essentially RX/TX out of it could be entirely PCIe-based, like the InfinityFabric stuff in AMDs CPUs. AMDs north/south-bridges are already PCIe-based anyways.

With such an approach, the chiplets/dies could actually be bonded straight to the CPU-board, unless an interposer makes sense, and the chiplets could be spread out more due to fewer space constraints (think ThreadRipper-big-CPU-modules).

>>105827310
Most of those are connections to DRAM. If RAM is integrated/soldered on the CPU-modules like it's done with GPUs, no need for that many pins. Alternatively, the RAM could move to the back of such a slotted CPU module, if slotted/upgradable RAM is still a requirement. If designed well, this would allow for trashcan-mac-style modular PCs with high performance and density.

>>105827371
that has been done. it's fucking stupid because at this point you're pretty much just doing a single board computer with a backplane for expansion cards.

>>105824454
Fucking retard, the heatsink had to be like that because the cache was external on either side of the CPU.

>>105827405
Yeah, which is kind of neat, though. Skip the ISA/PCI/PCIe-slot-shape and make a design that allows for any custom cooling solution on top, no integrated rear ports, and the slot could be standardized for allowing very wild case designs. Small form factor all the way up to 19" backplane-kind-of systems? Ditch the limits of ATX and variants.

>>105827558
Also, since VRMs and RAM could be soldered, there could easily be essentially universal waterblocks cooling all of the CPU-module's components, instead of mainboard-specific waterblocks for the same purpose.

>>105827558
>>105827565
that's still just a single board computer. how would this be any better than a fucking mini itx board

>>105824454
>see? It only needs a heatsink twice the size of the equivalent K6-2/K6-3 for the same TDP

You forgot that the CPU was also 2-3x the performance of a K6-2 (the K6-3 was so extremely rare it doesn't really count, especially since Athlons were already out by then)

>>105827709
>I'm retarded

>>105827652
A mITX board doesn't have any way to get more than PCIe x16 out of it.

I miss slotkets. No bent pins.

And if you need more pins just double the channels, it's not that hard.

>>105828250
m.2

>>105828606
The wider (physical) the bus the harder it is to keep frequency high and sync. That's why ATA was abandoned in favour of SATA, or SCSI for SAS

from a marketing standpoint it felt and looked premium compared to amd, it was also easier to install

ill give you a good one: hybrid HDD/SSDs

those were doomed to fail

>>105827991
>retard
except the P2 annihilates the K6-2 in actual floating point tests, hell it even gets beaten by faggot Celery

>>105827371
>Most of those are connections to DRAM.
Not even a third.