Linus Torvalds Rips Into Intel

ecktt said:
Windows! Linux is not all bad but you'd swear they were trying to be.

and your partly right, I don't like user experience software for Linux. I pretty much hate Linux monolithic kernel and having to program for back in school. Man, even updating drivers is a pain. Linus attitude over the years has not helped. The man all but picked a fight with Tanenbaum, the dude who wrote the book on Operating Systems. You just can't study Computers at a reputable university and avoid reading his book.

Intel screwed the pooch with meltdown but all CPUs with branch prediction are exposed to spectre. Him being a Jackhole about isn't helping anyone. And well I just have a pet peeve about the excessive use of the "literally"

:)
Click to expand...
You "literally" haven't used Linux lately, it seems.

And tossing Linus under the bus is like tossing God under the bus for those pesky pedopriests. Not to mention Linus has the chops that you distinctly do not.
 
Madoc said:
You know, if I have to explain my posts, I feel like I've failed somehow. :/

That post was a paraphrase from a scene in the movie Heavy Metal. It sprang into my head when I thought about how POed Torvalds seemed with Intel. Hopefully *some* people got the joke. Now I'll just go sit in the corner for a while and think about what I've done.
Click to expand...

"Hannover Fiste"!

Damn, I loved that movie...
 
OrangeKhrush said:
What? you mean to tell me a company with the good ethics of Intel is trying to use this as a means to get people to upgrade.......how dare you.

But yeah it seems like Intel is in IDGAF mode about older PC's. Haswell should still be fine but i can see Intel borking it. but I will not be buying intel this round
Click to expand...
It's the old Wintel hegemony raising its planned obsolescence head again...
 
SvenBent said:
Because ale of those OEM machines that runs with just barely enough cooling will suddenly run with not enough cooling cripling performance even more as they every so often gets downscaled to 1600mhz.
Also see: Dell and the first computer with the pentium4 800mhz busspeed hyperthreaded cpu running almost constantly at throttled speed.

You personal CPU example is a small drop in the oceans of computers out there.
Click to expand...

who seriously needs more than 1 core a turbo speed.. but for those that do and obviously operate in a datacenter where cooling isn't an issue, or have sufficient cooling -- why 'impose' a reduction of performance without offering a 'sweetener'... 'my' CPU is a drop in the ocean but of all the CPU's Intel sell's the OEM box is a drop in the ocean for them.. let them throttle but give the customer a 'chance' not just impose a reduction an everything ;)
 
Last edited:
Seems like there is an undercurrent to not let AMD show up intel over this.
 
n31l said:
who seriously needs more than 1 core a turbo speed.. but for those that do and obviously operate in a datacenter where cooling isn't an issue, or have sufficient cooling -- why 'impose' a reduction of performance without offering a 'sweetener'... 'my' CPU is a drop in the ocean but of all the CPU's Intel sell's the OEM box is a drop in the ocean for them.. let them throttle but give the customer a 'chance' not just impose a reduction an everything ;)
Click to expand...

You must have a hard time picturering outside the box of your own usage if you really believe in the above.
There is no sweetner comming for double performance drops (reduce performance from slow kernel access + throtling CPU speed.
and can you clarify how you believe intel OEM cpu'er are just a drop in the ocean I am not sure you know how many CPU goes to the OEM market but if you do have number to clarify that this is just drop in the ocean please enlightment me.
 
SvenBent said:
You must have a hard time picturering outside the box of your own usage if you really believe in the above.
There is no sweetner comming for double performance drops (reduce performance from slow kernel access + throtling CPU speed.
and can you clarify how you believe intel OEM cpu'er are just a drop in the ocean I am not sure you know how many CPU goes to the OEM market but if you do have number to clarify that this is just drop in the ocean please enlightment me.
Click to expand...

The point of my argument is Intel CPU's have loads of headroom, they can unlock all the cores or more than just allowing turbo on a couple of cores as a 'sweetner'.. yeah I know they wont but they 'should'. If the user is a heavy user, folding or bitcoin, then the benefit might be slight with basic coolers but they wont have standard OEM cooling, even if they do let the CPU throttle itself and give the user the option.

I've been overclocking Intel CPU's for years and even overclocked they never really hit thermal limits.. eg my EVGA SR-2 had x5690's overclocked to 4.5ghz (+50%) and I was folding on them 24x7 and they had no issue even using fairly basic EVO 212 coolers.

I used to work for what was the largest OEM manufacture and the users that bought hardware that might actually see a hit i.e. server users, I can assure you they have adequate cooling to unlock the cores and never see a throttle.. the 'home user' part of their business might not have the best cooling, granted, but they also wont really see the hit everyone is fussing about and again why not let the CPU control it's self, the 10 seconds a 'home user' might use 3-4 cores at max turbo probably wont even raise the temp enough to cause concern.
 
Last edited:
n31l said:
The point of my argument is Intel CPU's have loads of headroom, they can unlock all the cores or more than just allowing turbo on a couple of cores as a 'sweetner'.. yeah I know they wont but they 'should'. If the user is a heavy user, folding or bitcoin, then the benefit might be slight with basic coolers but they wont have standard OEM cooling, even if they do let the CPU throttle itself and give the user the option.

I've been overclocking Intel CPU's for years and even overclocked they never really hit thermal limits..
Click to expand...

and I am stil lsaying you suggestion is clearly horrible for anyone grasping anykind of company risc assetsment. If you think intel will do this and risc millions of computer going into throttling mode becacse oem machines that are bult with absolut minimum cooling can't handle the extra heat from this.
That you are constalty trying to argure with your system only clearly make you unable to understand that the million and millions of oem computer are not built like yours. Why this is hard to understand I dont know I have alrey giving you clear examples on the issue hapenning before.

Again if you only thinkg folding at how and bitcoin mining uses more than 1 core you are again highly misunderstadning differnet peoples workload OEM machine or no. and that is probalby why you dont get why your suggestion is not in anyway a good one on a company scale.


In short. The world is not you.
Intel has to think of a bigger picture than just whatyou deal with


P.S.
I asked you for evidence that OEM macihnes are just a small drop in the ocean of intels CPU's
you have not providede that.
 
SvenBent said:
and I am stil lsaying you suggestion is clearly horrible for anyone grasping anykind of company risc assetsment. If you think intel will do this and risc millions of computer going into throttling mode becacse oem machines that are bult with absolut minimum cooling can't handle the extra heat from this.
That you are constalty trying to argure with your system only clearly make you unable to understand that the million and millions of oem computer are not built like yours. Why this is hard to understand I dont know I have alrey giving you clear examples on the issue hapenning before.

Again if you only thinkg folding at how and bitcoin mining uses more than 1 core you are again highly misunderstadning differnet peoples workload OEM machine or no. and that is probalby why you dont get why your suggestion is not in anyway a good one on a company scale.


In short. The world is not you.
Intel has to think of a bigger picture than just whatyou deal with


P.S.
I asked you for evidence that OEM macihnes are just a small drop in the ocean of intels CPU's
you have not providede that.
Click to expand...

Your reading what you want to read.. 'systems' thats plural..

Throttling doesn't need to mean underperformance, just throttle down to the same limits the CPU's has currently but if thermals allow let all cores run free... and giving the user the 'option' if its too risky don't enable it - I never said force it upon people.

Do you really think an OEM makes an individual CPU cooler for every CPU type or one that covers a broad spectrum? (and before you start, yes I know they make more than 1 heatsink but not 1 for every SKU that your implying)

Do I really need to list every type of software on Windows & Linux just so you can't find something to get your knickers in a twist over? They were examples of apps that would use all cores and for prolonged periods to highlight even then max temperatures will never be seen on 'servers'.. your sig, seems to imply you have experience with 'heavy multitasking' -- load your servers up and I bet there's 20-30c headroom to reach intel's recommended maximum.

P.S. I don't care if you 'asked' you should learn to ask nicely, even then I'm bound by a non-disclosure agreement... so go find someone else to argue with.. you sound like a right jumped up little Herbert.
 
OrangeKhrush mvmiller12 Vegas P11 kac77 DeathFromBelow, The reason why I wasn't posting in this thread is because the key message was posted in #23, much before you guys started mentioning me. :whistle:

David explains why Linus is wrong. David mentions AMD and mentions Intel, because this issue affects to both companies. David explains why the Retpoline patch developed by Google isn't enough, and the IBRS patch developed by Intel is needed (at least temporally) to cover those holes not covered by Retpoline. David mentions compilers. He does because the retpoline approach requires a retpoline-enhanced compiler; this means that adding retpoline patch to the OS kernel isn't enough because older applications or any recent application compiled with normal compiler are still open to attacks. One of the reasons why both AMD and Intel will be changing the silicon soon.
 
juanrga said:
OrangeKhrush mvmiller12 Vegas P11 kac77 DeathFromBelow, The reason why I wasn't posting in this thread is because the key message was posted in #23, much before you guys started mentioning me. :whistle:

David explains why Linus is wrong. David mentions AMD and mentions Intel, because this issue affects to both companies. David explains why the Retpoline patch developed by Google isn't enough, and the IBRS patch developed by Intel is needed (at least temporally) to cover those holes not covered by Retpoline. David mentions compilers. He does because the retpoline approach requires a retpoline-enhanced compiler; this means that adding retpoline patch to the OS kernel isn't enough because older applications or any recent application compiled with normal compiler are still open to attacks. One of the reasons why both AMD and Intel will be changing the silicon soon.
Click to expand...
This is pretty much known in the linux world and hence why SpectreV1 mitigation is only now being merged into Kernel-.4.16. NOTE: kernel-4.15 was only released last week and equally gcc-7.3 was only released 2weeks ago BOTH are needed for mitigation.

Unsure in the windows world but all I can see is there has been not alot of patches for something very serious
 
naib said:
This is pretty much known in the linux world and hence why SpectreV1 mitigation is only now being merged into Kernel-.4.16. NOTE: kernel-4.15 was only released last week and equally gcc-7.3 was only released 2weeks ago BOTH are needed for mitigation.

Unsure in the windows world but all I can see is there has been not alot of patches for something very serious
Click to expand...

Spectre patches are already incorporated in the Windows kernel. People is awaiting AMD to issue the needed microcode to activate the v2 patch.
 
juanrga said:
Spectre patches are already incorporated in the Windows kernel. People is awaiting AMD to issue the needed microcode to activate the v2 patch.
Click to expand...
thats my point though... Spectre cannot just be protected by the kernel it needs a compiler update and thus where is the update to everything .. antivirus exe, calc.exe EVERYTHING
 
naib said:
thats my point though... Spectre cannot just be protected by the kernel it needs a compiler update and thus where is the update to everything .. antivirus exe, calc.exe EVERYTHING
Click to expand...

Reason why we need silicon changes.
 
You must log in or register to reply here.
Back
Top