Is there a performance database?

[Lord_Humungous]

BTW Complex33, what the hell is 'yaring'?

[Digital Concepts]

Interesting spreadsheet - nice work.

Add more FFT ranges, and more color codes - why I'd nominate ya for a Humungous award!

[Xyzzy]

I sure wish I had Excel... Can anyone convert it to a PDF or something?

[Stormblade]

Email on the way Xyzzy

[Stormblade]

Quoting myself from above...

As for the amount of memory used, of course the more you use, the better chance you have of finding a factor during P1 stage 2 testing and saving yourself a lot of time.. I think current doublechecks (8M range) will use 2-300 MB for optimum checking. Dunno about first time LL or 33M exponents.. Anyone care to help out on these?

Apparently I was a little wrong on the above. As long as you have enough ram for P1 stage 2 to run, then it doesnt matter too much...

Quote George from another forum.... (empasis mine)

The affect of addidional ram on P-1 is way too hard too explain.

From memory (so I may be a little off), stage 2 requires 4 temporaries, but would like to have 12 temporary variables. Each temporary for a 1MB FFT size will take a little over 8 MB. If you don't give prime95 enough memory to allocate these 12 temporaries, then prime95 will have to make multiple passes and thus be slower.

If you give prime95 enough memory to allocate more than 12 temporaries, then prime95 can do a few more optimizations to run a tiny bit faster. However, prime95 needs several temps more than 12 to use these optimizations. This is why you'll often see prime95 not use all the memory you give it. If it can't allocate 8 (or whatever the number is) extra temporaries it will just allocate the 12.

We're not done yet. The bounds picking code also notes how many temporaries can be allocated and will adjust the selected bounds slightly - deeper B1 and shallower B2 for low memory machines, shallower B1 and deeper B2 for more memory. These changes affect your percentage chance by only a very tiny bit - maybe a few tenths of one percent. Oddly, the fact that you have more memory and P-1 runs faster MAY cause prime95 to pick different bounds such that you spend more time on P-1 factoring than an identical machine with less memory.

More memory will have a marginally better chance of finding factors with more bits. However, once again, the effect will be too small for you to notice.

In summary, extra memory has a very, very minor effect on your chances of finding a P-1 factor. I certainly would not go out of my way to buy extra memory solely for prime95 to use.

[ebx]

CPU clock isnt the only factor. L2 cache size/speed, memory type,
speed will all make a difference. A 166DDR AMD XP 1800+ beats
a 133DDR 2100+ by a big margin.

On the other hand, if you look closely to the numbers for the fast
Celeron using a P4 core, you will find the big performance gap
between 786K and 896K FFT table sizes. Why? The only explanation
is that Celeron has only 128K cache. I would guess the 256K limit
will show when we hit 2000K tables. Would the new beta test that
size?

The newer NorthWood P4 is so much nicer that it mutes AMD not
only for pure performance but also for price/performance. I dont
have the hard numbers but it should consume less power and generate
less heat too. All count for gimps. Go for it!

[Complex33]

Yaring is the age old practice of making up nonsense in the attept to confuse and distract others. In other words I'm just babbling becuase I had to have some interest ;)

Complex

[Xyzzy]

So if I understand this right, the ideal setup is to have enough memory for the 12 temporary things... Anything more is really just a waste...

How can I know how much memory I need to support these 12 temporary things?

The reason I ask is because right now I run P-1 testing exclusively... I have 800MB assigned to it, but I'd like to assign less so I can do other things if I want to... Is there a formula? I'm in the 448K FFT range right now...

So far I have 5 factors found in 58 exponents, which is more than was predicted by Prime95 (4.63%)...

[Joe O]

Complex33,
What version of Excel are you using? I have 7.0a (for W95) and can't open it
Joe O.

[Prime95]

Originally Posted by Xyzzy

So if I understand this right, the ideal setup is to have enough memory for the 12 temporary things... Anything more is really just a waste...

How can I know how much memory I need to support these 12 temporary things?

From the readme.txt:

Exponent Minimum Reasonable Desirable
-------- ------- ---------- ---------
6000000 12MB 23MB 33MB
10000000 19MB 36MB 53MB
33000000 65MB 125MB 185MB

You want the desirable column. 53MB will more than cover your situation. Prime95 will put the extra memory you give it to some use - it wont waste it, so just set your memory to whatever you are comfortable with.

[binarydigits]

Don't toss those K6s just yet - I find they're great for trial factoring.