• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.

Linus Torvalds Upgrades to AMD Threadripper After 15 Years with Intel

Joined
Feb 19, 2009
Messages
1,162 (0.20/day)
Location
I live in Norway
Processor R9 5800x3d | R7 3900X | 4800H | 2x Xeon gold 6142
Motherboard Asrock X570M | AB350M Pro 4 | Asus Tuf A15
Cooling Air | Air | duh laptop
Memory 64gb G.skill SniperX @3600 CL16 | 128gb | 32GB | 192gb
Video Card(s) RTX 4080 |Quadro P5000 | RTX2060M
Storage Many drives
Display(s) AW3423dwf.
Case Jonsbo D41
Power Supply Corsair RM850x
Mouse g502 Lightspeed
Keyboard G913 tkl
Software win11, proxmox
I mean, it's the best tool for the job, no doubt about it.

I understand him skipping 1st gen Ryzen (they really had unfixable Linux issues), but now there is no excuse.

I like my 8-core Intel but if I could afford a threadripper, I'd be all over that. Believe it or not, I'm only on this because someone else got out and I got it on a deal. It's good for the money I spent, but far better exists and I know it.

which unfixable issues, segfault errors wasn't a linux issue, it was a broken cpu and you could rma it, my 1700 is working perfectly in linux on my server box.
it's just that.. it wasn't the best but hell at 329 at launch for an 8 core 16 thread it was who doesn't want one for such uses ?
Zen+ still wasn't the best other than TR platform which was superior and zen2 it's a no question kinda deal.
 
Joined
Aug 20, 2007
Messages
21,529 (3.40/day)
System Name Pioneer
Processor Ryzen R9 9950X
Motherboard GIGABYTE Aorus Elite X670 AX
Cooling Noctua NH-D15 + A whole lotta Sunon and Corsair Maglev blower fans...
Memory 64GB (4x 16GB) G.Skill Flare X5 @ DDR5-6000 CL30
Video Card(s) XFX RX 7900 XTX Speedster Merc 310
Storage Intel 905p Optane 960GB boot, +2x Crucial P5 Plus 2TB PCIe 4.0 NVMe SSDs
Display(s) 55" LG 55" B9 OLED 4K Display
Case Thermaltake Core X31
Audio Device(s) TOSLINK->Schiit Modi MB->Asgard 2 DAC Amp->AKG Pro K712 Headphones or HDMI->B9 OLED
Power Supply FSP Hydro Ti Pro 850W
Mouse Logitech G305 Lightspeed Wireless
Keyboard WASD Code v3 with Cherry Green keyswitches + PBT DS keycaps
Software Gentoo Linux x64 / Windows 11 Enterprise IoT 2024
which unfixable issues, segfault errors wasn't a linux issue, it was a broken cpu and you could rma it, my 1700 is working perfectly in linux on my server box.

Yeah, it was early die spinnings and I meant unfixable in software, of course you could RMA it for a new spinning that fixed it(I did and they overnighted me a new 1800X), for most that is fine as an option if you get a bad seed but for a linux kernel dev downtime is the devil, hence my comment.

The real issue was AMD never made a new product number for this spinning, hence it's a guessing game.

Knew it had to be something other than what PSP originally meant.

It "originally" meant Progressive Supranuclear Palsy, an advanced bad neurodegenerative disease. We are all stealing from people with rotting nervous systems.

 

ARF

Joined
Jan 28, 2020
Messages
4,670 (2.61/day)
Location
Ex-usa | slava the trolls
At least he did finally upgrade with something 3 times faster, even despite it being from the enemy camp :D

15 years with Intel systems, looks very suspicious and anti-logical.
 
Joined
Mar 31, 2020
Messages
50 (0.03/day)
Linus is smart enough to do his own calculations (if he had the time). He had his reasons and made his choice, much of our commenting have been a waste of time (here and on other forums).

If you really want an explanation/speculation ...
1) Compiling with multiple threads has diminishing returns when increasing CPU cores as you hit portions of the process waiting for 1 or 2 threads left to finish. Disk I/O can now keep up with 3x CPU performance when comparing NVMe SSD to SATA SSD. The whole system being new, it's hard to tell what impact each component has on the result. Compiling can involve lots of branch instructions, cache size and small IO.

2) Linus spends most of the year with emails and looking at git/code where it doesn't need a lot of CPU. Then it comes to the merge window and then for several days it's lot's of compiling and testing making use of the extra CPU cores.

3) Linus doesn't upgrade every 2 years, more like 4 years. Intel have been recycling 14nm designs for about 5 years and making minor improvements then offered factory overclocks. After getting the kernel compile time from over 65secs down to about 25secs, an extra 5secs is not as significant and not worth spending $1000's more. His computer at home is probably sitting idle while travelling, phone calls and during meetings.

4) Linus was frustrated by Intel's response to Spectre and Meltdown and other security vulnerabilities may have influenced the decision. Linus in general doesn't like bugs, poor security, slow responses, lack of information, crude fixes that greatly impact performance (eg. disable HT they said) but I'm not pointing the finger at just Intel, has happened to many. Linus was not rushing to AMD when past AMD based systems had MB/BIOS bugs and poor/late drivers for Linux (CPU/APU/GPU). Now AMD systems and software have proved stable, mature and worth the chance.

5) A 32 core AMD is maybe 2.5x to 3x the performance of an Intel 10 or 8-core from the last 4 years ago. The 32 core AMD is double the compile performance of Intel_14_core, 1.7x Intel_18_cores while saving 30% power at peak loads. But, the total system power used by the AMD TR is slightly more than Intel under low loads. Both get very hot when using all-core turbo to the max (AMD = less peak W per core).

et. al.
 
Last edited:

ARF

Joined
Jan 28, 2020
Messages
4,670 (2.61/day)
Location
Ex-usa | slava the trolls
1) Compiling with multiple threads has diminishing returns when increasing CPU cores as you hit portions of the process waiting for 1 or 2 threads left to finish. Disk I/O can now keep up with 3x CPU performance when comparing NVMe SSD to SATA SSD. The whole system being new, it's hard to tell what impact each component has on the result. Compiling can involve lots of branch instructions, cache size and small IO.


What about optimisations of the software for AMD's Zen architecture if he didn't touch any AMD CPU, to begin with.

And people are wondering why AMD has so poor support.


It's just too many shills and unprofessionalism.
 
Top