Talk:Mining hardware comparison

Question about mega-hashes per Watt versus mega-hashes per Joule. Does anyone have an opinion? Most people know what a Watt is but I think seeing MHash/J could be confusing to some people. While perhaps being more accurate, using "/J" might go over some people's heads, especially since the edit was given with little to no explanation. Vast 06:45, 19 April 2011 (GMT)

How about adding a note about what a Joule is somewhere people will see? --TiagoTiago 21:40, 9 May 2011 (GMT)

Hmmmmm... if only there was an online encyclopedia people could look it up in... Physics 06:52, 13 May 2011 (GMT)

In my opinion the MHash/s/W was more enlighting for non physicans, but hey, I don't really care. SmokeTooMuch 22:12, 30 May 2011 (GMT)

Is the note about the AMD Stream SDK valid anymore? Looking down the list it seems people are doing just as well and sometimes better on the 2.3 and 2.4 SDK as compared to the 2.1. Tybeet 23:04, 5 May 2011 (GMT)

It depends on the mining application and the OpenCL kernel used I think. Phoenix with phatk mines faster on SDK 2.4 than on SDK 2.1 under 32-bit Windows 7. Icaci 18:27, 29 May 2011 (GMT)

Color Coding

What is the color coding indicative of?

I think it is just a way to easily group cards of each type Fnord123 15:44, 23 May 2011 (GMT)

Power Calculation

How do people calculate power? I took the idle power numbers from Tom's Hardware, then measured incremental power w/a Kill-A-Watt when hashing on my ATI 6970 (its the one w/1.96 Mhashes/W). Fnord123 15:42, 23 May 2011 (GMT)

OS Listing

Think extracting out an OS column would be useful? Possibly OS version or architecture as well.

Cheapest card per mhash/s ?

i'll pick ... the 5830 --Compn 16:51, 3 June 2011 (GMT)

Point of reference or benchmark chart?

This page could use some serious cleaning up.

I think it needs to be decided if this page is a point of reference or a collection of benchmarks. If it's a point of reference, there should only be one entry per hardware type (video card model, cpu, etc.) with stock performance. Some video card entries have over six variations with different over clocking/settings.

Identical (or near-identical) entries aren't that useful, but other than I find it helpful to see what variations worked, even for a given card, especially with so many variables in play. Mumpsimus 01:17, 15 June 2011 (GMT)

Clarification of model number vs. number of cards.

Jthibo 18:10, 15 June 2011 (GMT)

I've noticed that people are putting in things like 6990x2 to indicate having 2, 6990 dual-gpu cards. The problem is that this is also how AMD numbers some of their card models to indicate the number of GPUs on the card. For example, I have a 4870 X2, which is the model number of the card (the X2 indicates dual-gpu, as opposed to the regular, single GPU 4870.) I see some people have put in entries for a 4870x2, but is this 2 (4870) cards, or 1 (4870 X2) card (or even 2 (4870 X2) cards?)

Perhaps we should have quantity number placed in parenthesis, so 2 4870 cards would be "4870 (x2)", and 1 4870 X2 card would be "4870 X2"? Or some variation on that. Perhaps even another column denoting the total number of GPUs for the entry?

This is a comparison of hardware, 2 * 5770, for example, is no more than useful than a single 5770, yet a card with 2 GPUs on, is useful. Multi-GPU setups should be removed, imo. --Rallan 19:22, 15 June 2011 (GMT)

Although this is not just a hardware comparison - software used, versions, parameters are very usuful here, but these x2 listings are confusing, and most importantly performance should be reported on a per GPU basis, i.e. divided by 2, with x2 or x4 in the notes column -- Vlnv 2011-06-24