Discussion the Death of the Desktop CPU

Page 4 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

gdansk

Diamond Member
Feb 8, 2011
4,040
6,658
136
Perhaps you'll get a server chip scaled down to desktop
What do you mean perhaps? That's been the case for like 8 years now.

You get server cores and you'll like it. And even if they switch to Halo-like multi-die module for desktop the Zen cores will still be a server core first and foremost.

It's an absurd thread. A 16 core multi-chip module with full-width AVX-512 and SMT is not a mobile first design. It's a hand-me-down of a hand-me-down from the market AMD actually cares about.

The actual mobile part - Strix Point - gets a core that makes so little sense for mobile that they chopped bits off, reduced buffers and added another core type and it's still a poor market fit.
 
Last edited:

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,986
3,440
126
Perhaps you'll get a server chip scaled down to desktop,

And what would you call this then?



A workstation CPU oh wait.. isnt that what you just said we perhaps would get?
Wouldnt my Threadripper be a scaled down EPYC?
 

johnsonwax

Member
Jun 27, 2024
140
227
76
And what would you call this then?

View attachment 121471

A workstation CPU oh wait.. isnt that what you just said we perhaps would get?
Wouldnt my Threadripper be a scaled down EPYC?
Why do you guys keep throwing that out as counter-evidence for what I said, when it is evidence for what I said? Is it a chip designed for desktop? Nope. It's a server chip scaled down. Desktop CPU is dead, all you will get is derivatives of chips designed for other markets shoehorned into your use, whether you like it or not. Lot of multicore at the expense of single core (which is why the top of the line AMD X3D is slower single core than an iPad Pro). Apple is running it the other way scaling up laptop to desktop. Microsoft is asking Qualcomm to do the same thing.

The point is, nobody is designing for desktop, and that's not about to change.
 

Cogman

Lifer
Sep 19, 2000
10,284
137
106
What will kill the desktop CPU if anything is DDR socketing. The fact is, you can cram in more memory with a lower latency and higher bandwidth if you ditch the DDR main spec and instead go with something like LPDDR5 (which requires soldered chips).

Further being able to setup a unified architecture with the CPU/GPU in one package and unified memory is really something that might take off in terms of what you can do.

I don't like this, I'd rather be able to swap in new ram, a new CPU, or a new GPU. However, that's simply where the manufacturers appear to want to be headed.

Here's to hoping CAMM2 pans out and maybe something similar to this for GPU socketing so that unified memory can be more of a thing without ditching upgradability.
 

Thunder 57

Diamond Member
Aug 19, 2007
3,438
5,675
136
What will kill the desktop CPU if anything is DDR socketing. The fact is, you can cram in more memory with a lower latency and higher bandwidth if you ditch the DDR main spec and instead go with something like LPDDR5 (which requires soldered chips).

Further being able to setup a unified architecture with the CPU/GPU in one package and unified memory is really something that might take off in terms of what you can do.

I don't like this, I'd rather be able to swap in new ram, a new CPU, or a new GPU. However, that's simply where the manufacturers appear to want to be headed.

Here's to hoping CAMM2 pans out and maybe something similar to this for GPU socketing so that unified memory can be more of a thing without ditching upgradability.

LPDDR latecny is poo for desktops/servers.
 

DaaQ

Golden Member
Dec 8, 2018
1,778
1,255
136
there are no "desktop parts" anymore... it's the same CPU and difference is very marginal



exactly. but some will go to great lengths defending how "superior desktop CPUs are", they don't realize it's the same

Well yeah but geekbench doesn't take cpus thermally throttling, or pretty much all the other bits of a computer that affects performance. I posted that benchmark showing my phone beating my desktop but I'd be very surprised if there were many real tasks where my phone was faster.
At least you can argue that a 3dmark benchmark is testing a metric that is closer to a situation that users are likely to run.
Exactly, let the above get NON RTX laptop get this. (I have no clue what RTX laptop chips score anyway)



For giggles


#77 is mine.
 

Hulk

Diamond Member
Oct 9, 1999
5,062
3,547
136
Ideally if I could have all of my applications and data in my phone, with sufficient local storage all I would need is a portable battery powered monitor and a desktop monitor, a keyboard and mouse.

Then when I get near the display the phone connects wirelessly to the display and I'm off and running.

But this is fantasy. I'd need more compute than is available on a phone, about 4TB of local storage, and all of my Windows applications.

Still, THAT would signal the end of not only the desktop PC for me but the laptop PC as well.

One device to rule them all.

Total convergence.
 

zir_blazer

Golden Member
Jun 6, 2013
1,214
506
136
Further being able to setup a unified architecture with the CPU/GPU in one package and unified memory is really something that might take off in terms of what you can do.

Where I have hear that before...


LPDDR latecny is poo for desktops/servers.

Since Cache purpose is to hide memory latency, it would be interesing if in Cache heavy SoCs like AMD 3DX the extra bandwidth can compensate for the worse latency. Perhaps it may be viable for big APUs, with LPDDR with tons of Cache. Isn't that how they do in consoles by using GDDR as shared memory, but actually lacking the big Cache?
 

aigomorla

CPU, Cases&Cooling Mod PC Gaming Mod Elite Member
Super Moderator
Sep 28, 2005
20,986
3,440
126
Perhaps you'll get a server chip scaled down to desktop,

Uhhh didn't you ask for this?


Why do you guys keep throwing that out as counter-evidence for what I said, when it is evidence for what I said? Is it a chip designed for desktop? Nope. It's a server chip scaled down.

Aren't you admitting to what i showed you being what you asked for? If you scale down a server chip doesn't that make it a desktop? I mean, not all servers need to be inside pizzaboxes with 9 million RPM 40mm fans that sound like a jet engine when they boot up.

OH just to note....
There was a EXPERIMENTAL INTEL server board which used 2 LAPTOP PROCESSORS called the INTEL SOSSAMAN.

It was intel's attempt at ultra low power servers, which later morphed into the entire XEON ATOM line.

But the earily designs were like this:


You can see the Laptop Sockets here:


And yes they also were fun to play with.
 
Last edited:
Reactions: Mopetar

Thunder 57

Diamond Member
Aug 19, 2007
3,438
5,675
136
Where I have hear that before...




Since Cache purpose is to hide memory latency, it would be interesing if in Cache heavy SoCs like AMD 3DX the extra bandwidth can compensate for the worse latency. Perhaps it may be viable for big APUs, with LPDDR with tons of Cache. Isn't that how they do in consoles by using GDDR as shared memory, but actually lacking the big Cache?

Whenever this topic comes up I like to search for this old review. Granted that's GDDR not LPDDR so latency is going to be way worse. You can see the trend though. CPU's like their low latency memory.
 

poke01

Diamond Member
Mar 8, 2022
3,393
4,637
106
if the 9950X was BGA but performed the same as an LGA 9950X in a desktop, can we consider it dead?

Do not confuse modularity as old or dead. In desktops and servers, parts should never be soldered.
 
Reactions: Thunder 57

WelshBloke

Lifer
Jan 12, 2005
32,376
10,487
136
Ideally if I could have all of my applications and data in my phone, with sufficient local storage all I would need is a portable battery powered monitor and a desktop monitor, a keyboard and mouse.

Then when I get near the display the phone connects wirelessly to the display and I'm off and running.
You've just described Samsung Dex.
 

johnsonwax

Member
Jun 27, 2024
140
227
76
Uhhh didn't you ask for this?
It wasn't a challenge, it was an expression, as in you will get either laptop scaled up or (perhaps) server scaled down, but you won't get a designed for desktop CPU. The point was, you don't get to choose.

And no, if you scale down a server processor to desktop, you're almost certainly sacrificing single-core, which is pretty important for a whole lot of things.
 

Thunder 57

Diamond Member
Aug 19, 2007
3,438
5,675
136
It wasn't a challenge, it was an expression, as in you will get either laptop scaled up or (perhaps) server scaled down, but you won't get a designed for desktop CPU. The point was, you don't get to choose.

And no, if you scale down a server processor to desktop, you're almost certainly sacrificing single-core, which is pretty important for a whole lot of things.

Nobody has designed a "Desktop CPU" in 20+ years. It doesn't make sense. I'm not sure what you are getting at other than being pedantic. It's far more cost effective to deisgn a good server CPU, and run many of them in a power/frequency optimized range. Then, take a version of that, juice the power/clocks and release it for desktop. It has worked quite well. What would your mythical "Desktop CPU" do better, and how?
 
Reactions: mikeymikec

mikeymikec

Lifer
May 19, 2011
19,784
13,795
136
The only ways I see the mainstream desktop dying are the following possibilities:

1 - A revolutionary step forward in processor design, e.g. superconducting at room temperature in an implementation that's practically available for mainstream use. Something so revolutionary that to try and scale it upwards to be bigger/better/faster is pointless in mainstream use and also impractically expensive for mainstream use.

2 - Microsoft successfully kills the mainstream desktop, e.g. with a successful move to ARM. Considering how Microsoft is imitating Apple at every turn in recent years, I see Microsoft attempting this as a distinct possibility.

I've used the word 'mainstream' because with the second possibility I still don't think the desktop form factor would be completely killed off, but it might be the end of many hardware manufacturers geared towards widespread desktop PC production. We might end up with a bunch of niche 'Silicon Graphics workstation' style computer makers for very expensive professional use.

Until then, whether a CPU manufacturer makes a mobile CPU that's derivative of a desktop CPU or vice versa and scaled respectively, desktop PC life goes on.
 
Reactions: Thunder 57

zir_blazer

Golden Member
Jun 6, 2013
1,214
506
136
Whenever this topic comes up I like to search for this old review. Granted that's GDDR not LPDDR so latency is going to be way worse. You can see the trend though. CPU's like their low latency memory.
I specifically mentioned a Cache heavy part to compensate for the higher memory latency. How much Cache L3 it has? 8 MiB for 8 cores? Doesn't fit the experiment.
Sadly, we don't have similar 3DX parts to compare it with.
 

Doug S

Diamond Member
Feb 8, 2020
3,094
5,335
136
What will kill the desktop CPU if anything is DDR socketing. The fact is, you can cram in more memory with a lower latency and higher bandwidth if you ditch the DDR main spec and instead go with something like LPDDR5 (which requires soldered chips).

Further being able to setup a unified architecture with the CPU/GPU in one package and unified memory is really something that might take off in terms of what you can do.

I don't like this, I'd rather be able to swap in new ram, a new CPU, or a new GPU. However, that's simply where the manufacturers appear to want to be headed.

Here's to hoping CAMM2 pans out and maybe something similar to this for GPU socketing so that unified memory can be more of a thing without ditching upgradability.

DDR6 appears likely to take up more board space because I believe DDR6 DIMMs will be 96 bits wide. I say that because rumor is that DDR6 will have four channels, which makes no sense for ECC if it stays 64 bits wide. However, with ECC becoming mandatory (or at least the bits needed for it becoming mandatory) in LPDDR6, which also moves from 16 bit to 24 bit width, it makes perfect sense if DDR6 channels are similar to LPDDR6's 24 bit channels.

Assuming LPCAMM2 becomes a thing, I think DDR6 will be mostly relegated to servers and high end workstations.
 

adroc_thurston

Diamond Member
Jul 2, 2023
5,396
7,569
96
The fact is, you can cram in more memory with a lower latency and higher bandwidth if you ditch the DDR main spec and instead go with something like LPDDR5 (which requires soldered chips).
a) there's like two different standards of mezzanine bolt-on LPDDR
b) it's way way way higher latency than the mainline DDR.
However, with ECC becoming mandatory (or at least the bits needed for it becoming mandatory) in LPDDR6, which also moves from 16 bit to 24 bit width, it makes perfect sense if DDR6 channels are similar to LPDDR6's 24 bit channels.
calling dat thing ECC is being very generous.
 
Reactions: Mopetar
sale-70-410-exam    | Exam-200-125-pdf    | we-sale-70-410-exam    | hot-sale-70-410-exam    | Latest-exam-700-603-Dumps    | Dumps-98-363-exams-date    | Certs-200-125-date    | Dumps-300-075-exams-date    | hot-sale-book-C8010-726-book    | Hot-Sale-200-310-Exam    | Exam-Description-200-310-dumps?    | hot-sale-book-200-125-book    | Latest-Updated-300-209-Exam    | Dumps-210-260-exams-date    | Download-200-125-Exam-PDF    | Exam-Description-300-101-dumps    | Certs-300-101-date    | Hot-Sale-300-075-Exam    | Latest-exam-200-125-Dumps    | Exam-Description-200-125-dumps    | Latest-Updated-300-075-Exam    | hot-sale-book-210-260-book    | Dumps-200-901-exams-date    | Certs-200-901-date    | Latest-exam-1Z0-062-Dumps    | Hot-Sale-1Z0-062-Exam    | Certs-CSSLP-date    | 100%-Pass-70-383-Exams    | Latest-JN0-360-real-exam-questions    | 100%-Pass-4A0-100-Real-Exam-Questions    | Dumps-300-135-exams-date    | Passed-200-105-Tech-Exams    | Latest-Updated-200-310-Exam    | Download-300-070-Exam-PDF    | Hot-Sale-JN0-360-Exam    | 100%-Pass-JN0-360-Exams    | 100%-Pass-JN0-360-Real-Exam-Questions    | Dumps-JN0-360-exams-date    | Exam-Description-1Z0-876-dumps    | Latest-exam-1Z0-876-Dumps    | Dumps-HPE0-Y53-exams-date    | 2017-Latest-HPE0-Y53-Exam    | 100%-Pass-HPE0-Y53-Real-Exam-Questions    | Pass-4A0-100-Exam    | Latest-4A0-100-Questions    | Dumps-98-365-exams-date    | 2017-Latest-98-365-Exam    | 100%-Pass-VCS-254-Exams    | 2017-Latest-VCS-273-Exam    | Dumps-200-355-exams-date    | 2017-Latest-300-320-Exam    | Pass-300-101-Exam    | 100%-Pass-300-115-Exams    |
http://www.portvapes.co.uk/    | http://www.portvapes.co.uk/    |