AMD or Intel as well as the memory brand - page 70

 
Mathemat >> :

228 * 3.8 = 866.4.

Wonderland: the efficiency increases with frequency and is now substantially higher than W5590. Can anyone explain it? Yura, didn't you make something else up?

Lyosha. No, I don't think so.

Docent >> :

Yes, interesting. And so far it's not very clear to me.

How was the overclocking performed?

What's inconvenient is that the test is not running under native OS and it's impossible to view CPU-Z with full information about the processor and memory.

The only thing is that calculation of t*f for Core ich makes no sense since only base frequency is taken into account while turbo boost is not. That said, it's quite possible that the memory is also overclocked (to DDR3-1600 type mode) and the Core i7-920 @3.8 + turbo boost will run faster than the Xeon @3.47 + turbo boost but with slower DDR2-1333.

The margin of error is not that big, and is probably determined by this.


I just tweaked the frequency - instructions are http://www.thg.ru/cpu/intel_core_i7_efficiency/onepage.html


tweeked only the CPU to 3.8 and left the DRAM voltage untouched

Base Frequency 133 MHz 150 MHz 166 MHz 183 MHz

190 MHz

200 MHz
CPU Frequency 2.67 GHz 3 GHz 3.33 GHz 3.66 GHz

3.8 GHz

4 GHz
CPU Voltage Auto Auto Auto 1275 В

1,344 В

1,5 В
GPL Auto Auto Auto 1,8 В

1,94 В

2,1 В
QPI/DRAM voltage Auto Auto Auto 1,275 В

1,344 В

1,5 В
DRAM voltage Auto Auto Auto 1,6 В 1,6 В 1,6 В

--


--

all other points were met

Paul's instructions https://forum.mql4.com/ru/25722/page68

excluding p5 and p6

--

I'll try to repeat it a few times and double-check everything

 

What kind of memory you have, you didn't say.

 
Mathemat >> :

What's your memory, you didn't say.

OCZ 3x1333lv6gk 12g memory, memory with undervoltage, ( but the OS is given 3 )

p6t ws pro motherboard

--


repeated the run with a time of 03:47

double-checked - seems correct

---

Just kidding: I'm taking orders for a quick optimization :-))))


--

p.s another run the tester gave the time 03:46

again everything was correct according to the manual

 
Docent >>: It is quite possible that the memory is also overclocked (to DDR3-1600 type mode) and the Core i7-920 @3.8 + turbo boost will run faster than the Xeon @3.47 + turbo boost but with slower DDR2-1333.

There is another hypothesis. At this frequency the FSB of the stone equals or becomes close to the memory frequency (they are in sync, while the memory frequency can stand still). The memory communication bus bandwidth increases dramatically. So it turns out that you raise frequency by 43% and performance goes up by 61%. It's a miracle!

 
Docent >> :

Yes, interesting. And so far it's not very clear to me.

How was the overclocking performed?

It's inconvenient that the test is not running under a native OS and you can't view CPU-Z with full information about the processor and memory.

The only thing is that calculation of t*f for Core ich makes no sense since only base frequency is taken into account while turbo boost is not. That said, it's quite possible that the memory is also overclocked (to DDR3-1600 type mode) and the Core i7-920 @3.8 + turbo boost will run faster than the Xeon @3.47 + turbo boost but with slower DDR2-1333.

The margin of error is not that big, and is probably determined by this.

3.47 is exactly the turbo boost. The frequency of my Heon is 3.33

 

 
Mathemat >> :

There is one more hypothesis. At this frequency the FSB of the stone is equalized or becomes close to the memory frequency (they are synchronized, while the memory frequency may stand still). The memory communication bus bandwidth increases dramatically. So it turns out that you raise frequency by 43% and performance goes up by 61%. Amazing!

There's no FSB in non-Halems - it's different there. But that's not the point - it's still kind of weird. The hell with all these parallels, virtual boxes and other virtual stuff. You will never get the truth from them.

Yura, can we just take some memory off the board and (for the umpteenth time!))) test it?

 

First test: 369 * 2.66 = 982

Second test, overclocking: 228 * 3.8 = 866.4

Third test, more overclocking: 214 * 4 = 856

The ff-performance continues to rise. Choo-de-sa!

P.S. I daresay that at 800 efficiency it will burn out (Nehalem's energy protection is sort of disabled?). Just kidding, of course.

 
Mathemat >> :

First test: 369 * 2.66 = 982

Second test, overclocking: 228 * 3.8 = 866.4

Third test, more overclocking: 214 * 4 = 856

The ff-performance continues to rise. Choo-de-sa!

P.S. I daresay it will burn out at 800 efficiency (Nehalem's energy protection is sort of disabled?). Just kidding, of course.

No! It will move to a new qualitative state like superconductivity! It will go into a state of supercalculation and take results directly from the world's information field!!! Oh, yeah!!!

 
Svinozavr >> :

There's no FSB in non-Halems - it's different there. But that's not the point - it's still kind of weird. The hell with all these parallels, virtual boxes and other virtual stuff. You will never get the truth from them.

Yura, and it is possible to take off a part of memory from board and (once again!))) to test it?

01) Manufacturer OCZ
02) Model OCZ3X1333LV6GK
03) LifeTime warranty
04) Frequency 1333
05) Type DDR3
06) Bandwidth 10666
07) Timings 7-7-7-20
08) Voltage 1.65V
09) Cooling Radiators

--

http://x-memory.ru/club/forum/forum1002/topic572/

my machine has 12 gig, 3 x 4 gig memory cards - i got a set of 3 ( look - 6 gig memory cards are available on the web site now... )

Honestly, I'm too lazy to take it off and open the case etc.

--

what is the possible effect of this procedure?

Besides, the operating system gets 3 gigs of memory.