• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

HyperThreading on or off?

newtekie1

Semi-Retired Folder
Joined
Nov 22, 2005
Messages
28,473 (4.00/day)
Location
Indiana, USA
Processor Intel Core i7 10850K@5.2GHz
Motherboard AsRock Z470 Taichi
Cooling Corsair H115i Pro w/ Noctua NF-A14 Fans
Memory 32GB DDR4-3600
Video Card(s) RTX 2070 Super
Storage 500GB SX8200 Pro + 8TB with 1TB SSD Cache
Display(s) Acer Nitro VG280K 4K 28"
Case Fractal Design Define S
Audio Device(s) Onboard is good enough for me
Power Supply eVGA SuperNOVA 1000w G3
Software Windows 10 Pro x64
Something I noticed in my IntelBurnTest compilation thread was that the i7/i5/i3 CPUs seem to crunch numbers a heck of a lot better with HyperThreading off, and IntelBurnTest is multithreaded.

So with bigadv work units, do they go faster with HT off also?
 
Keep it on. Multithreaded performance dropped by over 60% with hyperthreading off in my testing (all cores loaded to 100%). The reason being is that 8 > 4. Each work unit gets done a little slower but it is working on twice as many simutaneously.
 
Keep it on. Multithreaded performance dropped by over 60% with hyperthreading off in my testing (all cores loaded to 100%). The reason being is that 8 > 4. Each work unit gets done a little slower but it is working on twice as many simutaneously.

Your comment doesn't make sense then. If each unit gets done a little slower, that means lower PPD, no?:confused:

Like I said, in my number crunching tests with IntelBurnTest to see the actual GFLOPS my i7 and i3 were capable of, turning Hyperthreading off improved performance. Reason being is that there is an inefficiency introduced with swapping out the registers and cache in the cores to switch between two threads.

I guess I'll test myself.
 
Depends on the program being used, it's a shame hyper threading isn't a dynamic process and only switches on when you need lots of threads.
 
Leave it on on the I series processors unless your overclock is causing to much HEAT.:toast:
 
Newtekie, I remember when I used to do SETI@Home pre-BOINC. I had an Athlon 3200+ (actually an o/c 2500+) and it would crunch units at a certain rate. Remember, these were single core 32-bit processors back then. <nostalgic moment>

I then got a P4 Northwood with HT. Run the same seti with HT off and it was a bit slower than the Athlon. But switch HT on and it would race past the Athlon at about twice the speed! It was awesome. :D

What happened, is that switching on HT slightly reduced the speed of each individual seti unit, but then it did two of them simultaneously, which more than made up for it. It's a similar principle to how nvidia's GTX295 was faster than a GTX285 by using slower cores in SLI mode to make it faster overall than the 285.

So, in your case, just see which setting gives you more completed work units in unit time and leave it there. Each case is different and the advantage with HT on or off depends on the details.
 
Anytime you are folding with a I3/I5/I7, leave hyperthreading on. Your folding does get accomplised quicker than if it was turned off.
 
Definately keep it on.
 
Definately keep it on.

+1. Leave it on. When folding -bigadv work units, the 4 physical & 4 virtual cores trick the work unit into thinking it is a dual-quad rig(8 cores). Otherwise, the work unit would not run for a normal quad. Core i7 rigs must be clocked to approx 3.6Ghz+ to meet the bonus deadline for -bigadv(which is substantial). Most of us run them @ 3.7 to 4.1ghz depending on thermal solutions for the CPU.
 
3.8GHz is supposed to be the sweet spot for i7s.
 
+1. Leave it on. When folding -bigadv work units, the 4 physical & 4 virtual cores trick the work unit into thinking it is a dual-quad rig(8 cores). Otherwise, the work unit would not run for a normal quad. Core i7 rigs must be clocked to approx 3.6Ghz+ to meet the bonus deadline for -bigadv(which is substantial). Most of us run them @ 3.7 to 4.1ghz depending on thermal solutions for the CPU.

I was hoping someone would have mentioned you can't run bigadv WUs on an i7 Quad without HT on :toast:
 
I then got a P4 Northwood with HT. Run the same seti with HT off and it was a bit slower than the Athlon. But switch HT on and it would race past the Athlon at about twice the speed! It was awesome. :D
I saw the same results with my benchmarking app. i7 w/o HT is a cripple. :(
 
Actually, you can run the bigadv wu's on I7's at 3.2 ghz and still get the bonus, and more if it is a 2685 wu.
 
Well I kicked off hyperthreading and the time per frame went up by 6 minutes, which was enough to make me loose the points bonus deadline.

+1. Leave it on. When folding -bigadv work units, the 4 physical & 4 virtual cores trick the work unit into thinking it is a dual-quad rig(8 cores). Otherwise, the work unit would not run for a normal quad. Core i7 rigs must be clocked to approx 3.6Ghz+ to meet the bonus deadline for -bigadv(which is substantial). Most of us run them @ 3.7 to 4.1ghz depending on thermal solutions for the CPU.

Mines at 3.6GHz, but I might bump it up, but my TRUE is already struggling at 3.6GHz, with temps in the 70-72°C range when folding.

I was hoping someone would have mentioned you can't run bigadv WUs on an i7 Quad without HT on :toast:

I didn't have a problem. I just had to have it download with HT on, then turn it off.
 
I leave mine on. then again i have i3
 
with f@h wcg etc etc as long as they are multi threaded apps you will get much higher PPD with HT ON. yes each wu goes about 30% slower but the number of wu's being crunched t one time makes up for and then some. i tested this out when i first got my i7 860. its much much more ppd
 
with f@h wcg etc etc as long as they are multi threaded apps you will get much higher PPD with HT ON. yes each wu goes about 30% slower but the number of wu's being crunched t one time makes up for and then some. i tested this out when i first got my i7 860. its much much more ppd

With F@H that is definitely not true. With SMP folding you are working on one WU at a time. But yes, the single WU does seem to go faster with HT on according to my testing, and the WU actually does complete faster.
 
I think OC101's comment about multiple WUs applies to WCG.
 
I think OC101's comment about multiple WUs applies to WCG.

Yeah, I know that is how WCG works, but I was just correcting the F@H part.:toast:
 
Definitely on ;)

Off, I get ~9-10k PPD @ 3,83ghz, on I can get 15-16k @ same clocks!
 
Back
Top