Sunday, 31 October 2010

Loot


I've got plenty of loot queued up to review, just had a bit of a hiatus for a while. The SSD has had time to become part of the collective, but was a definite 'aha moment' for the technologist in me.

...And it went something like this. The shift between spending time and spending money - A Good HDD enclosure costs about two-thirds the price of an SSD drive, and then you have to buy the HDD too. The end result is that the SSD is cheaper, fast, quieter, lighter, and physically smaller than its magnetic uncle at the cost of lower storage capacity.

So on the subject of terrabytes, What needeth thou? You want some local storage, which is now a cache/buffer between me and the internet - the advantage of a cache being its close by and fast but capacity has increased faster than data rate such that a modern drive is 'slower' than ever. After my Level1, I've RAM, virtual RAM, local storage, NAS and then then internets with each layer getting more bloated, harder to search, slower and further away.

Anyway I'm here to talk about loot. The Latest loot being the SSD drive with nowhere to mount inside my case - which obviously means the next loot added to the basket will be a new case. Or two new cases, as the aha moment of storage was a mirror of my build philosophy which - of late - has been to put everything in a box and hope it boots. So I've found the excuse to talk about build strategy which only adds to the list of unblogged topics I want to pen and havn't gotten around to reviewinf the 64GB Crucial SSD that is lying forlornly on the bottom of my case suspended at one end by a power cable that has to navigate the passive cooler of an oversized video.

More about loot after I have sobered up. I really want some of those socks though. You know the ones? Yeah. Those Ones.






Monday, 4 October 2010

Minecraft Rendering


So, Minecraft is getting great press. Whether it a flash-in-the-pan boom and bust economy or a slow builder is irrelevant at this time. The Minecraft Guy had a pretty good idea, implemented it and is getting the recognition he deserves.

Pulling apart the game, landscapes seem pretty important. Without being able to render a vast landscape the game wouldn’t be worth a fraction of what it is. So I’ve been pondering on what it takes to pull off such a techincal feat.

The world is made of blocks. About 1m each is a fair estimation for a sense of scale, and the visible world extends to about 192 units until you reach the far clip plane. The far clip plane can be reduced to about 128 units by adding fog which provide us the ‘Far’ and ‘Normal’ draw distances and it can be reduced further yet with Short and Tiny draw distances.

192 blocks radius is a 384x384 area containing a theoretical 384x384x384 blocks in the observable world reaching a staggering 56Million blocks to cull and render.

A broadphase pass that doesn’t even attempt to render air and creates a render list based on the occlusion of block using its six neighbours drops these unimaginable millions to a contemplatable two dimensional 384x384 ‘surface’ approaching 150,000 cubes at six quads each, still getting on for a million quads.

150,000 is a lot of cubes, and view frustrum culling with a <= 90’ field of view will reduce this to a ball-park of (192*192) 36864 Cubes and CPU side backface culling gives you a three quad per cube total of 110,000 Quads. Its a lot, to be sure, but its a managable amount compared to the 300Million+ Quads of the unculled estimate.

110,000 Quads feels like a lot, but might just be enough to brute force. At this point I’d probably still assume we are Render Bound so a sensible step would be to divide the task separate threads for Render, Occlusion, Game and Network and rely on the Host OS to manage four threads well enough.

A closer far clip plane and a 128 unit draw distance does reduce this to around 50,000 Quads and a 64 Unit draw distance reduces it further to a mere 12,000 Quads to render. The draw distance can probably be increased above 192 quite cheaply with a combination of untextured blocks, chunk sub-sampling and render proxies.


Tuesday, 21 September 2010

Church Mouse


The latest addition to the family is a new JSCO noiseless gaming mouse. It may not look like much but its got it where it counts... Well, it doesn't have any balls but thats expected in this day and age. The JSCO mouse has REALLY REALLY quiet buttons. Quiet like you can hardly hear them, and it doesn't take long to really fall in love with its clickless slickness.

With four settings from 750-1600 dpi and a button to switch between them its darn quick to switch from a sensible desktop speed to an uber twitch fps turn-on-a-dime speed. Sounds like a gimmick, and probably is, but its a gimmick thats very likeable very quickly.

Two quiet thumb buttons, two quiet mousy buttons and ... Its Achilles wheel. The mouse wheel clicks like a Morse code dolphin. Actually clicking it is alright, but scrolling it just lets the whole package down. Feels like the wheel has left, right and center clicks giving you seven buttons plus uppy/downy scrolling which again is really nice.

Its a small mouse, lets call it compact. Smaller than the 'average' Microsoft mouse - you know the one - it kind of gets lost in the hand. But its not the size that counts, its what you do with it and you can do a lot with the JSCO without being heard.

Its a good mouse gaining high marks for excellent motion and a bunch of buttons. A couple of bonus points for being quiet and having dpi selection too gives this an 8/10. A quiet wheel and larger form factor would have earned it more points.






Thursday, 16 September 2010

Kernel Mustard


About 80GB into a backup, mah file server rebooted.
Do Not Want. This is bad, really bad if you value data security - which I do. So this is bad.

Applying some perspective this is something like 2000% better now its got a Promise SATA controller, which is a start. And I believe in starts. But I really want to be able to do whole machine backup-restore, as well as large scale copies of media.
Largly, because I've got a media centre PC and want to copy all of my music onto it. Worse than this, I've got a CD collection ripped to 320kbps MP3 that I've now got the storage space to rip to PCM. The magnitude of the task means I only want to do it once, and really want to rinse and repeat with my DVDs for that 21st Century On-Demand feeling. Physical scratchable disks, what is this - the nineties?

So an unstable fileserver is just NOT on the cards. FreeNAS 6.0 has been reliable for years but running on a low spec machine with 500GB bursting at the seams and this new(er) hardware and storage space should be enough to last for years to come.

So I'm about to go back to the log file(s) and find something new to diagnose and fix. All in all, its a bit of a bother. The project came about because I made some assumptions about reliability and stability which I'm still not sure about.

Recycling a 3.2Ghz Intel P4 with a few gigs of RAM as a fileserver seemed like a good idea, and seems like it should be stable enough. Its a known-good PC that served well as a desktop/gaming computer when it was up to the spec, and since being replaced putting some serious disks in it sounded like it would give it a new lease of life. RaidZ seemed like a good idea, FreeNAS seemed like a good idea.

So how come all of these good ideas haven't combined to form an uber-idea? The sum of the parts isn't all that grand and no matter how I look at it I can't help wondering if I'm using the wrongs tools for the job. What should be a really good domestic file server has tecked up until its become a second rate enterprise solution, and may have suffered in the process.

Setting my goalposts closer and storming past them might just serve better than aiming high and having to put effort in to reach them.








Saturday, 11 September 2010

The Promise Land

The continuing adventures of Alt-F4

With the rising popularity of Microsoft, personal computers have entered the mainstream and all but replaced the abacus and etch-a-sketch that they resemble. However the technologist community have had a mixed opinion of MS offerings and viable alternatives have been quietly queuing up for attention. There are a lot of reasons to take what you are given and like it without questioning what you are getting, everything is simpler. I'd be happy that a computer works in the same way that a car does. I have a vague idea what makes it go, and I understand that there are people who do - but I don't need to be one of them to be happy.
But then you can raise your head above the trenches and take a good look around. What is new, what works, what can I achieve? What can I do to push myself, what can I do to push the technology around me?

Apple are seeing a return to glory, clinging on to the mainframe age with their thin-client Pods and Pads and whatnot which provide a user interface to a real computer somewhere that handles All Your Stuff. Almost under the security radar because of an historically small user base, Apples consumer devices are more at risk from physical theft than virus attacks and probably will remain so for some time. Heads are nodded to their white plasticy appliance computing and its user-first paradigm.

Unix famously is still in the Data Centre game, with Sun Solaris breaking new ground with ZFS which provides a number of advantages over traditional RAID-5 solutions. This alone shows the company is alive on the cutting edge of its field, and data hosting is a billion dollar market worldwide.

Linux in all its forms is pushing to be desktop contender, but marginal games and application support put Linux third in a two-horse race for the mainstream market despite a low price point. Good flexibility and security make it more viable in the media-centre and netbook market. Despite being a tremendous community driven OS with all of the virtues of Open Source development, a lack of direction has kept Linux in the shadows and publicity like Xandros, and ChromeOS could significantly raise public awareness.

Back from theory to reality - This leaves me with a Windows desktop PC because I play games. An Ubuntu-Linux media centre, a BSD-ZFS file server and a Google-Linux netbook. I'll keep my apples in the fruit bowl for now. While it all seems like an unlikely mix, each decision has been made on its own merits and I've had the freedom to make that decision myself.

Technology integration being what it is, the compatibility problems of lifestyle are still too common and too painful for my liking. An entire Microsoft or Entire Apple setup would probably be a little smoother to troubleshoot but the lack of choice and control in both worlds is a little off-putting. At the moment I can solve my Windows problems with a reformat/reinstall, while the *nix problems generally take a kernel patch and a bit of googling.
The latest such problem has been reboots/dropouts and timeouts on my file server. The machine is an old Intel P4 running FreeNAS 7 with ZFS providing me a four drive RAID-Z pool of around 4.0TB. Its the sort of problem that makes you want to buy a prepackaged solution, however the domestic fileserver market hasn't really picked up to my needs and enterprise offerings start and budgets I can't justify.

Google hinted that the problem could actually be the Silicon Image 3512 ATA controller (and drivers) rather than a disk problem or an issue with BSD, so I've replaced the controller with a Primise FastTrak TX4310 controller and all of my problems have gone away. With the SiI controller it took a lot of luck to copy more than a couple of GB without fail, but the first impressions of the Promise card are that its delivering.
The Final Score?
Silicon Image NILL - Promise Technology ONE.

And that concludes it. Another problem vaguely hopefully black-boxed. Find problem, find solution and apply one to the other until it all works.

It was a £100 investment for the controller, but an otherwise recycled PC has life for a few more years and now I have a lot more faith in the Promise Land of unlimited* local storage with world class redundancy to keep my data safe.

* 4TB isn't unlimited local storage, but its quite a bit.




Monday, 6 September 2010

On the Blog


I'm back on the blog. It sounds like a medical condition, and probably is, but I'm here to review the Silicon Image 3512 SATA controller in a mix of compliments and complaints.

The context is I recycled an old PC to use as a fileserver. Its an old Asus motherboard with an Intel 3.2Ghz HT-P4 and a few gigs of ram - nothing special by todays standards but probably capable of running a fileserver for domestic use - just has to store some music, photos, and backups of homebrew projects and code snippits. That sort of thing. Ideally, it'd have a full backup of all of my CDs and DVDs because shelf-space is at a premium and storage space is cheap.

BSD. FreeNAS. ZFS. 4x1.5TB RaidZ. CIFS. FTP.

Problem was the aged mobo only had two SATA ports. Leaping before I looked I rushed to the store and bought a two port SATA controller. Enter the SiI-3512.

After a number of problems, a kernel update, mailing list archives, forums and FAQs... I stumbled across a poster that said he's having trouble with his SiI-3512 SATA controller - and a lot of replies saying the same. Many of these with surprisingly similar error logs to my own, followed by a number of "Problem solved. Bought a Promise-SATA controller"

So, the compliments of the SiI-3512. It was cheap and store bought. You can't beat the convenience of going to a shop and coming out with your product in hand. It feels very real, involves human interaction, and is usually quicker than ordering online.

Complaints about the SiI-3512... Well, none yet. We'll just have to wait for the delivery time of the internet-ordered Promise controller - and enough time to install and configure the machine to use it.


Saturday, 8 May 2010

Axe to Grind

While its easy to get caught up in star ratings and numerical scores one can sit back and look at the porpoise of writing reviews at all. A wonderful specimin of cetacean, the porpoise will dive up to... hold on. That was a typo. I shall wax here about the purpose of reviews. I'm trying to scale each item on its own qualities by determining the scale on which it can be measured and placing it on that scale. How friendly to Dolphins can fishing for Tuna actually be, I mean, really? Friendly is the wrong scale of measurement. Fishing for Tuna can be measured on a dolphin scale of Killy, but killing less of them hardly categorises the process as friendly. One wonders if there are Tuna friendly Dolphins, which is unlikely.

Anyway, irrelevant as that was, what matters is the scale of measurements. Its very difficult to use an arbitary star rating largely because its arbitary, but its impossible to compare apples and oranges fairly until a scale and datum have been established. To this end, I'm going to reserve some space in each review explaining the comparisons but still try to avoid "This X is better that Y" statements.

So on the subject of grinding, I should probably mention the Krups grinder thats been serving me well. Its a Krups GVX2:
Which I bought because it satisfied all of my criteria, and it does exactly what it says on the tin. It'll grind the right quantity of coffee, producing even results and its fast to use.

Technically, for some processes you might want more than seventeen settings for the size of the grind but these only come into play under humidity, temperature and pressure conditions that arn't all that common at a fixed altitude in a temperate climate. If these become an issue you can counter a lot of the effects by switching to a moka pot rather than drip or press preparation and let physics do the hard work for you.

Likewise, quantity control is quantised to about 1/4 cup accuracy (1/8th of proper portions) but thats enough accuracy for my palette at least and fundementally it grinds all the beans you put in the top so if you put the right amount in you'll get the right amount out.

Its fast. Maybe a tad noisy, but I'm simply not measuring acoustics so have no complaint. It does take electricity, and lacks the contemplative satisfaction of a hand grinder due to its speed and the fact that you are no longer grinding the beans yourself. You have to decide when that is an issue and when it isn't. It also won't grind unless you've inserted the bin to catch the grinds and the hopper lid. Great for zero mess grinding but when those microswitches break it'll kaput the whole thing.

The down points a few and far between, but its a little difficult to clean. The exhaust nozzle that provides grinds gets a bit packed - especially with the finer grinds. When you switch to a coarse grind you can get up to a teaspoon of fine grinds and dust - very bad if you are switching from a paper microfilter to a cafetiere. This won't be a problem for day-to-day usage, where you may have a similar grind but cleaning can be a little fiddly.

While the cleaning process is a necessary evil, it is both an evil and necessary but the only frown on a so far faultless product.

Krups GVX2 Grinder - 9/10.