Quote from: bcp19 on August 04, 2013, 03:14:27 PM

Quote from: oxfeeefeee on August 04, 2013, 04:49:16 AM

Quote from: bcp19 on August 04, 2013, 04:28:36 AM

Quote from: oxfeeefeee on August 04, 2013, 03:31:11 AM

Quote from: bcp19 on August 04, 2013, 02:53:32 AM

Depends on speed. We know sieving is running about 25x, but how much of the CPU time was used for sieving and how much for fermat testing? Let x = CPU sieve time and y = CPU Fermat time. If x=y (50/50 split) then GPU = 2% for 52% of normal time = 92.3% speed increase. If x=9y (90/10 split) then GPU = 3.6% for 13.6% of normal time = 635% speed increase. if 9x=y (10/90 split) GPU = .4% fo 90.4% of normal time = 10.6% speed increase. These are extremely rough figures, communication times between GPU/CPU will also apply. Maybe he'll solve Fermat testing on the GPU and get a full 2500% speed increase. Short answer: too many current unknowns.

I did a similar calculation, but then Koooooj@reddit pointed out that I was wrong. Because with a much powerful sieve, CPU gets less test to do for fixed number of output.

for example (with made up numbers):
before: 10000Numbers ---sieve--> 100Numbers ---test---> 1Number
after : 10000Numbers ---sieve--> 10Numbers ---test---> 1Number

this way, even we don't speed up the execution of Fermat test, the test time is down to 10%.

But he's comparing apples to oranges. If sieveing speed is increased 25x, then you need different parameters to sieve deeper. Also, where does he get these numbers? 10000 seems extremely small, 100 is 1%, so we are to believe the CPU sieve is 99% effective? The GPU sieve is 99.9% effective? There's a lot of questions that need to be asked here:
What is he using for h₀?
What is he using for c?
What is he using for e?
What is the first sieving prime?
What is he using for P_s?

I'm sorry, but those number he's tossing out look made up, with no proof to back them up.

And I made up those number just to explain the idea, not him.
This is just another possibility, GPU could still be as slow as 2x

Ah yes, the old scientific "I pulled these numbers out of my ass" theorum. Great, thanks for playing... NOT!

On another note, he is comparing one gpu chip on a 6990 to one core on a Phenom II X6 1055T. The 1055T is a relatively weak processor and has 6 cores. So that 25x improvement turns into 25/6 = 4.2. Dividing again by 1.5 because the 1055T is weak we get 4.2/1.5 = 2.8. Not bad for this early in the game but also not a huge game changer either.

As bcp19 has been saying it also depends a lot on how much difference a fast sieve makes and how effectively the overall algorithm can make use of the gpu sieving. From what I have seen from posts by Sunny and Mikaelh it sounds like sieving is in the ballpark of 50% of the work. Only time will tell how it plays out on the gpu.

Regardless of all that if mtrly releases a working gpu miner that can match or beat an i7 3770 in the next week I will be impressed.

The 1055T compared to some other chips:

http://www.cpubenchmark.net/cpu.php?cpu=AMD+Phenom+II+X6+1055T