Bitcoin Forum

Bitcoin => Bitcoin Discussion => Topic started by: NewLibertyStandard on April 05, 2010, 08:20:01 AM



Title: Benchmark Utility
Post by: NewLibertyStandard on April 05, 2010, 08:20:01 AM
I'm testing a four core computer with hyper-threading and after one day of testing, which admittedly is an extremely small amount, it has only produced twice as many bitcoins as the slightly slower GHz dual core computer. When generating new blocks takes so long, it's hard to calculate performance. It would be useful if there was a stand alone utility or perhaps a utility built into Bitcoin which would state how many bitcoin generation attempts are being performed per minute. Then people would be able to change settings on their computer and after just a few minutes, verify how much of an affect those changes have. For example, I would like to see how much of a difference having hyper-threading turned on or off has on the amount of Bitcoin generation attempts occurring. I would also like to see how many more calculations per minute one computer has as opposed to the other.


Title: Re: Benchmark Utility
Post by: NewLibertyStandard on April 06, 2010, 04:04:45 AM
I decided that hyper-threading is better even if it were a bit slower because I want to keep one thread unused by Bitcoin and leaving a hyper-threaded half core unused by Bitcoin is better than leaving a whole core unused by Bitcoin. So scratch that request for a benchmark utility. There are other more important features to work on for now.


Title: Re: Benchmark Utility
Post by: gkelly on April 10, 2010, 05:50:09 PM
I decided that hyper-threading is better even if it were a bit slower because I want to keep one thread unused by Bitcoin and leaving a hyper-threaded half core unused by Bitcoin is better than leaving a whole core unused by Bitcoin. So scratch that request for a benchmark utility. There are other more important features to work on for now.

I think that a very useful thing to have in the BitCoin main window would be an indicator saying "# hashes/sec" to demonstrate how fast the BitcoinMiner threads were running.


Title: Re: Benchmark Utility
Post by: theymos on April 14, 2010, 01:51:35 AM
If BitCoin counted the number of hashes it calculates per second/minute, it would be possible to determine the average number of blocks that will be solved per day. This would be very useful.

If hashing works the way I think it does (nonce is incremented per try), this data could be gathered efficiently by saving the start nonce and start time and comparing it to the end nonce and end time whenever the nonce is regenerated.


Title: Re: Benchmark Utility
Post by: BitcoinFX on April 15, 2010, 03:58:39 AM
If BitCoin counted the number of hashes it calculates per second/minute, it would be possible to determine the average number of blocks that will be solved per day. This would be very useful.

Yes. It would be very useful and it will become somewhat essential in calculating a more accurate exchange rate in the future !

Regardless of the number of individual user connections, I would also like to see the total number of connected nodes vs the number of 'generating' nodes in the Bitcoin interface. It would also be good to know the highest and lowest number of connected nodes in any 24hr period.

As Bitcoins usage becomes more 'Macro', just calculating daily exchange rates is not going to be enough to satisfy Bitcoins diverse 'digital' community. Using 'Micro' exchange rate calculations from this data and the existing factors that NLS already includes will enable many more possibilities such a 'live trading'.

It would also be interesting to see the total number of transactions and the total amount of those transactions, again for a 24hr period.

You don't need to affect any aspects relating to user anonymity. You just need to harness the readily available data to plot high, low, medium and the mode.

What I'm basically saying is that this is a 'digital' currency (obviously), yet the projects 'base rate' is still effectively being calculated using 'paper' math !!!

This is not to say, that NLS isn't already doing a great job ! Not to mention satoshi's ( and any others ) excellent work in building the thing.

I think I will make a request list and add a topic in Technical...

I'm far from an 'expert' programmer, but with much IT and trading experience I have a good idea of what 'could' be calculated and added for the benefit of future users.

Having glanced over the 'active' source code much of the above is already being used in the 'functioning' of the software. I don't know whats planned for future releases, but we must have the foresight to collect and harness this type of data. Can any of the above be 'historically' viewed via irc data ? I've been making backwards neural network calculations and future predictions and would like to continue with this to help the project.