Bitcoin Forum
December 08, 2016, 08:15:28 PM *
News: To be able to use the next phase of the beta forum software, please ensure that your email address is correct/functional.
 
   Home   Help Search Donate Login Register  
Pages: [1]
  Print  
Author Topic: The determining factor on MH/s?  (Read 2749 times)
grue
Global Moderator
Legendary
*
Offline Offline

Activity: 1932



View Profile
March 12, 2011, 05:54:24 PM
 #1

What's the most important factor that determines a card's hashing ability? From what i heard, it's based on the amount of shaders. is that correct? what about core clock?

It is pitch black. You are likely to be eaten by a grue.

Tired of annoying signature ads? Ad block for signatures
1481228128
Hero Member
*
Offline Offline

Posts: 1481228128

View Profile Personal Message (Offline)

Ignore
1481228128
Reply with quote  #2

1481228128
Report to moderator
Advertised sites are not endorsed by the Bitcoin Forum. They may be unsafe, untrustworthy, or illegal in your jurisdiction. Advertise here.
1481228128
Hero Member
*
Offline Offline

Posts: 1481228128

View Profile Personal Message (Offline)

Ignore
1481228128
Reply with quote  #2

1481228128
Report to moderator
1481228128
Hero Member
*
Offline Offline

Posts: 1481228128

View Profile Personal Message (Offline)

Ignore
1481228128
Reply with quote  #2

1481228128
Report to moderator
Dude65535
Full Member
***
Offline Offline

Activity: 126


View Profile
March 12, 2011, 06:01:39 PM
 #2

shaders, clock speed and chip design are the determining factors

1DCj8ZwGZXQqQhgv6eUEnWgsxo8BTMj3mT
Cryptoman
Hero Member
*****
Offline Offline

Activity: 728



View Profile
March 12, 2011, 06:06:28 PM
 #3

On a related note, does anyone know how many integer operations are necessary to compute one hash?

"A small body of determined spirits fired by an unquenchable faith in their mission can alter the course of history." --Gandhi
os008
Member
**
Offline Offline

Activity: 79



View Profile
March 12, 2011, 06:44:37 PM
 #4

Interesting question. I made a quick calculation ...

5850 -> 300 MH/s -> 2.5 TFLOPs

2,500,000,000 / 300,000,000 = 8.33 OPs

I'd say 9 OPs. Someone here who has more expertise in this, please provide some feedback.

2x5850 @ 600[M|K]H/s
Beremat
Member
**
Offline Offline

Activity: 118


View Profile
March 12, 2011, 06:52:43 PM
 #5

Interesting question. I made a quick calculation ...

5850 -> 300 MH/s -> 2.5 TFLOPs

2,500,000,000 / 300,000,000 = 8.33 OPs

I'd say 9 OPs. Someone here who has more expertise in this, please provide some feedback.
4850 -> 75MH/s -> 1TFLOP

1,000,000,000 / 75,000,000 = 13.333

It's probably dependent on core architecture as well.
Meni Rosenfeld
Donator
Legendary
*
Offline Offline

Activity: 1890



View Profile WWW
March 12, 2011, 07:11:22 PM
 #6

FLOPs are floating point operations per second, so I don't think they're what we're supposed to look at.

1EofoZNBhWQ3kxfKnvWkhtMns4AivZArhr   |   Who am I?   |   bitcoin-otc WoT
Bitcoil - Exchange bitcoins for ILS (thread)   |   Israel Bitcoin community homepage (thread)
Analysis of Bitcoin Pooled Mining Reward Systems (thread, summary)  |   PureMining - Infinite-term, deterministic mining bond
ArtForz
Sr. Member
****
Offline Offline

Activity: 406


View Profile
March 12, 2011, 07:13:47 PM
 #7

off-by-3-orders-of-magnitude.
HD5xxx needs about 4150 integer operations per hash
HD4xxx and nvidia 2/4/5xx about 6350 integer operations.
for all current architectures integer ops/second = number of shader ALUs * clock.
so a 5870 at 1600 ALUs * 850MHz = 1360000M intops/second, / 4150 = ~ 327.7Mh/s
4870: 800 * 750M / 6350 = ~ 94.5 Mh/s
GTX580: 512 * 1544M / 6350 = ~ 124.5Mh/s
etc...
usually published FLOPS numbers are simply intops * 2 (one muladd counts as 2 operations).

bitcoin: 1Fb77Xq5ePFER8GtKRn2KDbDTVpJKfKmpz
i0coin: jNdvyvd6v6gV3kVJLD7HsB5ZwHyHwAkfdw
os008
Member
**
Offline Offline

Activity: 79



View Profile
March 12, 2011, 07:29:54 PM
 #8

HD5xxx needs about 4150 integer operations per hash
Thank you for the explanation.

Quote
usually published FLOPS numbers are simply intops * 2 (one muladd counts as 2 operations).
Wouldn't that mean ATi is much superior to nVidia in FLOPs; contrary to the reality? Don't you mean FLOPs = INTOPs / 2 ?

2x5850 @ 600[M|K]H/s
Garrett Burgwardt
Sr. Member
****
Offline Offline

Activity: 350



View Profile
March 12, 2011, 07:38:04 PM
 #9

FLOPS are completely irrelevant, since the hashing algorithm uses fixed point calculations.

You could not be more wrong by looking at the FLOPS rating of a card.
os008
Member
**
Offline Offline

Activity: 79



View Profile
March 12, 2011, 08:14:36 PM
 #10

There is still a correlation between FLOPS and Integer operations.
That's what i would like to know; what's the equation to convert between them?

2x5850 @ 600[M|K]H/s
ArtForz
Sr. Member
****
Offline Offline

Activity: 406


View Profile
March 12, 2011, 08:27:49 PM
 #11

published FLOPS for GTX580: 1581.1G
512 ALUs * 1544 MHz shader clock * 2 (muladd factor) = 1581.056G
published FLOPS for HD5970: 4640G
3200 ALUs * 725MHz * 2 (muladd factor) = 4640G

hey, what a coincidence!

bitcoin: 1Fb77Xq5ePFER8GtKRn2KDbDTVpJKfKmpz
i0coin: jNdvyvd6v6gV3kVJLD7HsB5ZwHyHwAkfdw
gigabytecoin
Sr. Member
****
Offline Offline

Activity: 280


View Profile
March 12, 2011, 08:57:13 PM
 #12

FLOPS are completely irrelevant, since the hashing algorithm uses fixed point calculations.

You could not be more wrong by looking at the FLOPS rating of a card.

Ok, so where are the FPOPS ratings?! Tongue
os008
Member
**
Offline Offline

Activity: 79



View Profile
March 12, 2011, 09:10:16 PM
 #13

hey, what a coincidence!
So to finish an integer operation, it needs TWO sub-operations? I would've thought floats were harder.

Ok, so where are the FPOPS ratings?! Tongue
http://pastebin.com/AvymGnMJ
http://en.wikipedia.org/wiki/Comparison_of_ATI_Graphics_Processing_Units#PCIe_.28HD_5xxx.29

2x5850 @ 600[M|K]H/s
apofis
Member
**
Offline Offline

Activity: 77



View Profile
March 12, 2011, 09:11:36 PM
 #14

Look here http://forum.boinc.ru/yaf_postst395_Spicok-vidieokart-dlia-kranchingha-v--MilkyWay-home.aspx
grue
Global Moderator
Legendary
*
Offline Offline

Activity: 1932



View Profile
March 14, 2011, 03:03:10 PM
 #15

thanks for all the input guys.

It is pitch black. You are likely to be eaten by a grue.

Tired of annoying signature ads? Ad block for signatures
Pages: [1]
  Print  
 
Jump to:  

Sponsored by , a Bitcoin-accepting VPN.
Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2006-2009, Simple Machines Valid XHTML 1.0! Valid CSS!