AMD wins!

[stj]

Intel has fucked everybody again!!
are you ready for a 50% performance drop? - your gonna get it once Hyper-threading is removed.
https://www.theregister.co.uk/2019/0...g_mitigations/

[diif]

Re: AMD wins!

In the real world it's about 10% with a maximum of 30% if it's the ideal workload.

[stj]

Re: AMD wins!

sure it is.
even browsers are multi-threaded and exploit HT and multi-core cpu's now.

[Topcat]

Re: AMD wins!

My CPU's are so old, I can care less.

[BigTroll]

Re: AMD wins!

Yep I feel like a winner everytime I use my AMD RYZEN!!

[diif]

Re: AMD wins!

sure it is.
even browsers are multi-threaded and exploit HT and multi-core cpu's now.

Not disagreeing things use HT but suggesting half will be lost is way off.

It's going to be more noticeable in data centres though that's more where the ideal workloads are seen.

[RJARRRPCGP]

Re: AMD wins!

Intel has fucked everybody again!!
are you ready for a 50% performance drop? - your gonna get it once Hyper-threading is removed.
https://www.theregister.co.uk/2019/0...g_mitigations/

Reminds me of the "Bush Jr." era, where some people had HT-phobia. IIRC, back in the Pentium 4 era, (they were AKA "Pee 4") some people were telling other people to disable hyper-threading, accusing hyper-threading of causing malfunctions, LOL

[BigTroll]

Re: AMD wins!

Reminds me of the "Bush Jr." era, where some people had HT-phobia. IIRC, back in the Pentium 4 era, (they were AKA "Pee 4") some people were telling other people to disable hyper-threading, accusing hyper-threading of causing malfunctions, LOL

Interesting, I know a few years back I was running a optiplex 280 with a geforce 100 OEM card and I had put in a 3.2ghz CedarMill pentium 4. Playing youtube and watching taskmanager it was hitting 100% cpu and causing the video to drop frames where it was running great a few days ago. Noticed that the CMOS battery went dead and hyperthreading got disabled, turned it back on and with hyperthreading enabled cpu usage went down to 80-85% on youtube and the video playback was smooth! this was in 2014 or so when trying to use a P4 with the modern internet was on its way out.

As a side note I usually always bought i5 cpus as I could never justify the price premium of the i7 just to get hyperthreading and a slightly higher clock. this thread makes me feel better about that decision.

[eccerr0r]

Re: AMD wins!

People are confusing threading (software concept) versus threading (hardware concept). Disabling hardware multithreading will not make software threading take twice as long to complete.

The speed penalty depends on the application and the OS scheduling. The 10-30% is about right from my experience though even this is optimistic. As regards to BT's experience, this is because of poorly written "multithreaded" OS/applications, usually because the scheduler improperly can't or won't relinquish control to other threads, and having hardware threading will see a huge apparent increase in throughput because now the crap OS/software is letting hardware run the other thread.

Of course having another core is much better than a SMT thread. But I have seen the actual throughput increase before and it is real and I use it when I can when throughput matters. The largest measurable increase I've seen was running one of those distributed clients from distributed.net which is completely CPU bound (no network or disk usage) and saw that 25%-odd throughput increase when using both threads on a P4. Of course each individual thread took significantly longer to complete, but there was a significant throughput increase.

Again on an ideal OS and properly written software, you will see very little speed loss with the disabling of hardware multithreading, especially when the software is waiting for user input most of the time. But bad software that's completely chomping the CPU will see a penalty.

[RJARRRPCGP]

Re: AMD wins!

software that's completely chomping the CPU will see a penalty.

Fun fact: Goldeneye chomped on the Nintendo 64 CPU! (And it probably does chomp on the CPU more than Perfect Dark!)

[stj]

Re: AMD wins!

Not disagreeing things use HT but suggesting half will be lost is way off.

It's going to be more noticeable in data centres though that's more where the ideal workloads are seen.

apple says 40%,
so probably similar for Linux.
although mine is on an AMD dual core.

[mariushm]

Re: AMD wins!

For people that are not up to date ... new exploit found for Intel computers ... again related to HT/SMT ... NOT present in AMD and ARM processors

They called it Zombieload

Wired: Meltdown Redux: Intel Flaw Lets Hackers Siphon Secrets from Millions of PCs - https://www.wired.com/story/intel-md...cution-buffer/

Bleeping Computer : New RIDL and Fallout Attacks Impact All Modern Intel CPUs - https://www.bleepingcomputer.com/new...rn-intel-cpus/

[ratdude747]

Re: AMD wins!

My CPU's are so old, I can care less.

Maybe not? I wonder if Netburst's HT has the same flaw? What about Nehalem/Westmere EP (both of those have HT as well)? I know you have both still in service (as I do I). Westmere and Nehalem are not on the microcode fix list (on the "no changes" list), Which would imply the issue sandy bridge and later (which is on the fix list)... but as old as they are, it could be a case of "nobody uses those, aww f*ck it". Dunno.

Fix list referenced attached, which came from this advisory from Intel:

https://www.intel.com/content/www/us...-sa-00233.html

----

I only have 3 potentially effected systems in service: One netburst (irwindale), one Westmere EP, and one ivy bridge (main personal laptop, a tricked out Latitude E6430). I don't plan on disabling HT... unless I see concrete proof that said platform is indeed affected.

[Topcat]

Re: AMD wins!

Maybe not? I wonder if Netburst's HT has the same flaw? What about Nehalem/Westmere EP (both of those have HT as well)? I know you have both still in service (as I do I). Westmere and Nehalem are not on the microcode fix list (on the "no changes" list), Which would imply the issue sandy bridge and later (which is on the fix list)... but as old as they are, it could be a case of "nobody uses those, aww f*ck it". Dunno.

Fix list referenced attached, which came from this advisory from Intel:

https://www.intel.com/content/www/us...-sa-00233.html

----

I only have 3 potentially effected systems in service: One netburst (irwindale), one Westmere EP, and one ivy bridge (main personal laptop, a tricked out Latitude E6430). I don't plan on disabling HT... unless I see concrete proof that said platform is indeed affected.

My Harpertowns don't support HT....hmmmm I still have a paxville system in operation and several westmere's though.... I never did bother with spectre patches though....all my westmere systems aren't accessible to or from the outside world.....

[lti]

Re: AMD wins!

The list of affected CPUs includes CPUs without HT.

I have two affected CPUs. One is a Core i5 without HT, and the other was always unusually slow in every application except benchmarks.

Sometimes I wonder what would happen if I bought a Ryzen instead of the i5. Would I like it?

[Dan81]

Re: AMD wins!

Meanwhile, I'm smiling with 2 i3 CPUs (i3 540 and i3 3220), both with HT

[RJARRRPCGP]

Re: AMD wins!

Intel has fucked everybody again!!
are you ready for a 50% performance drop? - your gonna get it once Hyper-threading is removed.
https://www.theregister.co.uk/2019/0...g_mitigations/

Uh, reminds me of the earlier Pentium I stuff, even when it apparently carried on until at least 1995 or real close to 1995. (the floating point bug)

But I don't think it's quite as bad as that. But still bull! The floating point bug, would definitely make people required to throw their processors away! IIRC, Intel said that the 133s and higher, don't have the hardware bug.

[stj]

Re: AMD wins!

that FP bug thing was the thing that set me against Intel,
i always used cyrix or amd when i could after that.

it wasnt the FP bug, it was the "go fuck yourself" type response from intel that was the decider.
i obviously wasnt wrong - they still have the same response to users!

[Topcat]

Re: AMD wins!

^
Cyrix.....ohh the memories....memories of horrible performance, instability, and simply not working....I think I'd take the FPU bug. AMD wasn't much better back in those days. Today OTOH, if I had to build a new system for myself, AMD would certainly be on the table of consideration (whereas in the past it wouldn't be).

[stj]

Re: AMD wins!

my cyrix systems were fine,
there was an old rule - not cpu specific.
the rule was: avoid pc-chips and other "cheap" chipsets like the plague!!
i remember they had lots of issues if you tried to use the then-new pci slots for example.

and amd k6-2 chips were great.