IBT Speed reading

Hi all, erm weird question or for me atleast, I use IBT (Intel Burn Test) to test my stability and when I disable HT I get a speed reading of 55 GFlops and each test takes 139 seconds on very high stress level, so I tought to myself hey why don't I activate HT again because I have it so why not use it, started IBT and it shows me a speed of 46 GFlops and it takes 161 seconds for each test how is this possible, Should not HT increase the calculation speed of the computer or is it only me that think this is really weird?

I think Hyperthreading doesn't do much for tight loop arithmetic calculations. It will be running 8 threads on 4 physical cores, so there will be context switching between the threads, which will eat up cycles and hence reduce the floating point operations rate.

Hyperthreading will help when threads can go idle waiting for i/o, which is not the case with IBT.

Oh, I always tougt it helped even if its not things that have to do with video and audio rendering etc, but then I can just keep it off then and save myself the heat and much much lower vcore I can have then since all I do is gaming, secondly is it a huge performance increase in having 3.8 vs 4Ghz or does it give so little increase in games its not worth it?

Andreas W said:

Oh, I always tougt it helped even if its not things that have to do with video and audio rendering etc, but then I can just keep it off then and save myself the heat and much much lower vcore I can have then since all I do is gaming, secondly is it a huge performance increase in having 3.8 vs 4Ghz or does it give so little increase in games its not worth it?

Each time you switch a thread it takes CPU cycles. It only helps when cores go idle because the running thread is waiting for I/O and another thread can be scheduled to run on the core.

200 MHz - not worth it IMO.

As I tought, just going to keep it at 21x181 then, since it require a whole lot less voltage, so if I want more fps im better of overclocking the gpu I guess

Andreas W said:

As I tought, just going to keep it at 21x181 then, since it require a whole lot less voltage, so if I want more fps im better of overclocking the gpu I guess

Yes, but even overclocking the GPU doesn't buy you much in FPS, only more GPU cores does.

Oh, so with my current 560 I can just leave it at default speed and lower the voltage, I can run it at 1.00v at its default 925mhz clock because it dont help that much as I understand?

That had been my experience with a 450. I just ordered a 560 Ti so I can soon tell you what I think there

I thought the 560 default speed was 820 Mhz though.

This is a superclocked version what I understood, and it for some reason cost about the same as the ones with 8xxMhz, mine is not the TI thou. This is my card WEBHALLEN.com - Asus Geforce GTX 560 1GB DirectCU II TOP (ENGTX560 DCII TOP/2DI/1GD5) - PCI-E / DVI / HDMI its in Swedish but there is google translate, and for some reason its cheaper then the ones with less core clock, the only one I understand why it cost more is the TI version which have 384 cores while mine have 334, but I would love to hear your conclusion after testing

A followup on this. I measured the performance of my new 560 Ti FPB at three OC levels:
Reference (no OC): Core 822, memory 2004
FPB (vendor OC) : Core 850, memory 2002
OC ( my OC) : Core 940, memory 2065 (just a random that did not produce any artifacts at same FPB voltage)

OC is 10% over FPB and 14% over reference

I used two benchmarks - the synthetic EVGA SCANNER benchmark and Unigine Heaven 3.0 which is based on a gaming graphics engine. Heaven was set at maximum settings. All ran 1920x1200 resolution and on a 2500k overclocked to 4.5 GHz.

Results:

EVGA Scanner:
Reference: 85 fps
FPB: 88 fps
OC: 97 fps (14% over reference, 10% over FPB - scales like the core clock)

Heaven (score/ave fps/max fps/min fps):
Reference: 570/22.56/53.5/8.3
FPB: 590/23.4/54.9/8.7
OC: 635/25.2/58.9/8.4 (11% over reference, 7.6% over FPB min fps is about constant)

So the synthetic score scales just like the OC, while the more real-world scale less than the OC. by about 3%, and the minimum fps on Heaven was the same for all.