Bitcoin Forum

Here is a new Excel Doc I'm working on
Tell me what you think.

Prices updated...
ROI Added...
Overclocked MH/s

UPDATED June 5th
https://i.imgur.com/I98WS.gif

https://i.imgur.com/VhgVh.gif


Donations: 18W5htHMBMEBYUSkEqqEzANbizJZp5Q1pu

Great work, carbonc!

had a calc error.  Its fixed now

Very nice! Can you add the 6850 and 6870 please?

I created a very similar spreadsheet.

I also created a second sheet with system build costs included ($125/card additional cost), and the 5850 still won out, though the 5870 was much closer than with just the cost of the cards included.  This was when the 5850 was at $140 though.

Very nice - add the 6870 too please :)

We need data for 6xxx budget series... (6670 etc)!

Also it would be cool to have this on Google docs, might even be possible to embed it in the Wiki like this then.

Another 2 columns that would be interesting: "cost after x days with y Cents/kWh running 24/7"(buying + electricity derived from Watts), "billions/trillions/quadrillions(?) of hashes checked in these x days"

I guess I could hack together something like that quickly...

Added

6950
6870
6850

I don't have MH/s data from that series.
Getting my data from here:

https://en.bitcoin.it/wiki/Mining_hardware_comparison (https://en.bitcoin.it/wiki/Mining_hardware_comparison)

very optimistic numbers for both 5850's, 5870, 68x0.
hardly many even overclocked fit that range.

I think these stats are fine for avg users who plan to run graphics card at basicly stock.

For instance, the hd5850 listed as 314mhash/s while with minimal overclocking every single hd5850 of mine do 350mhash/s+

I agree about the average users but every card will perform as expected with these results on average
Overclocking only degrades the cards and shortens its life and many card manufactures will probaby vary with results.

I would presume that overclocking each card will give a certian percentage increase across the board.
This atleast gives users a baseline to start from...

How about adding another column on the spreadsheet to take into considering costs when overclocking with a modest 15% performance increase.

Would give a realistic result for those interested in buying something with overclocking taken into consideration.

300 MH/s on a 6870 is only possible with heavy overclocking imho, I run at ~990 MHz (already overclocked!) and have ~280 MH/s!
Another ~10% more as in this Excel sheet is quite hard to manage...

Ok, 5970 vs 6990 is wrong in W

5970 needs 294w if you don't try overclocking, right!
But the 6990 needs "just" 375w in same case.

I don't know there this data came from, but if its from here "https://en.bitcoin.it/wiki/Mining_hardware_comparison" you have to research all data if they are clocked or normal - if they are real live or manufacturer informations.

Next: 188w for the 5870

Ok, I'll do some more research.
Add in Stock vs Overclock seen by users... maybe an average overclock instead of max...?

Not everyone will be able to get the max MH/s some other user has posted.

Sorry for the typos on some of the readings.

I havn't built in a direct data pull from that wiki yet.
I'd like it to be real time with prices and user's MH/s results.

Maybe I would be better suited building a database with user entry capability.

Great post!

I'm just starting out... currently have 1 4870 and 1 4670 running with a 6870 and 5770 to replace those in the mail.  I'm wanting to gauge 3-4 weeks worth of results with these cards before moving forward with additional builds.

You are right. overclocking depends on your cooling. (and your nuts)  ;D

Next problem:    :P
The whole thing becomes really difficult if you try to calculate a rig with 3 or 4 cards.
You need to install different power supplies...
You perhaps can overclock four 5870, but whats about four 6990 with the same cooling..?

I tried to answer this, but my calculation and prices are in € and wouldn't fit here.

Here's a (imho) more useful representation (click on picture to see bigger version):

https://i.imgur.com/DnYJyl.jpg (http://imgur.com/DnYJy)

Y axis:
Total Megahashes (MH/s*60*60*24*[#ofdays]) until that day divided by EUR spent until that day (calculated from Wattage + costs per kWh)
X axis:
days

Here you can see that if you just plan to mine short term, a 5830 can be a VERY useful investment --> in the longer run however a 5970 might be hard to beat.

It's also quite fun to play around with electricity costs or MH/s for example and see the lines jump + rearrange! :)

It might also be useful to determine if you want to spend for example 500€ in the next 90 days to determine which card would bring the most performance until then.
(Multiple cards btw. have exactly the same ratios - 2 cards cost twice as much, use twice as much electricity and also calculate twice as many hashes)

Problem 1:
You can't calculate the hardware and power together, because hadware is something you can sell again.

Problem 2:
where ist the 5870? (best w/h card in my calc)

1: As long as you're mining, you can't sell the hardware. In the end you'll have to mine (converted to EUR/USD/whatever) about as much as the "costs after x days" column says minus what you could sell your hardware for (as there is no predefined value, I didn't include that). I also didn't include difficulty increases + projected income, as this is far too speculative.
It would be up to you to guess how much a used card might be worth after x days.

2: Just add it yourself?! It's not rocket science and I did this sheet in ~15 minutes.
If you want to have it from me, I could also sell it - 1 work hour @ 20EUR means this sheet is worth 1 BTC. PM me for an address.

Updated Today

Awesome! Glad I just purchased some 5830's :)

Please keep this updated as this is an awesome reference!

I will try.

Like an idiot, it bought 2 5830's on Thursday with 3 Day select shipping for free. 
I lost alot of potential BTC generation over the weekend because they are currently sitting in a terminal. 
Those lost 3 days would have paid for the next day air costs.  lol...

According to these numbers (and at ~20 cents/kWh) up to ~ 1 year 5830s are the best cards, if you want to mine longer, 5870s take over until ~1 1/2 years and if you want to mine even longer than that, 5970s are the best cards you can get.

I won't publish my sheet again unless someone really needs to see it though.

For shorter term mining up to one year btw. there is nothing even close to 5830s (if the prices, wattages and hashrates from the OP are correct!)

5970's are the best cards but now as of today they are going for $1000 USED on ebay.  I know people can sometimes find them on buyitnow auctions for less.  Those often last minutes before they are snapped up.

http://cgi.ebay.com/XFX-ATI-Radeon-HD-5970-HD597ACNF9-GDDR5-SDRAM-PCI-/150612846079?pt=PCC_Video_TV_Cards&hash=item231139a5ff

At 1000USD/piece they are below average at best... and get beaten for more than a YEAR by 2/3 of the cards on that list! They just catch up with a 6870 after 3 years...

Even though eBay charges now a whopping 10% fee, this might be a good point now for some people to start selling 5970s and (if they want to continue mining) buying 5870s/5830s and BTC directly.
Most newies will not do these calculations and get trapped deeply and quickly in some VERY bad deals with this, as they most likely only look at "MAX HASHRATE = MAX INCOME!".

I pay $.10/kwh so these cards are absolute steal at that price. Sadly, mine are also in transit.  :(

This means your 5830 cards are better than any other card for ~570 days hashes solved per USD invested wise. :)

However this does not mean that you will mine the most BTC! If the system keeps growing, the higher initial burst you can get, the better. At nearly any price/cost.

Are you going to Update this list sometime in the Future??

I like how the chart is setup however I see a couple problems


First... the 5970 w/ 800mhash?? You'd have to overclock a bit - or use flags That I dont know about because I have one and I get 600mhash at stock speeds w/ -v -w128. However - If there is a flag I should be running to get 800 mhash Id really like to know.

Secondly, the 6950s are a little fucked up, in a general sense. The older 6950s that were reference design could have their bios's flashed to unlock the shaders and perform as a 6970 does. This is technically buying a 6950 and turning it into a 6970. The new 6950s are not reference design - and only get 300 mhash because most of the cards cant unlock the extra shaders.

I'd also like to see the links to the specific cards, because there are many... are these manufacturer overclocked versions or what???