THE BOOK--Playing The Percentages In Baseball

What are you going to do with the box that will be CPU bound?

Are you buying a laptop or a desktop?  If you’re buying a laptop, you almost have to go with Intel right now.  AMD just can’t compete on battery life.  If you’re going desktop, AMD is pretty good especially for the dollars per performance.  AMD can’t compete yet with the Core i7 processors yet though.  In Q3 they are launching Llano which should be better than the Core i7 Sandybridge from Intel.  Mainly because the graphics core they are putting into the chip is much better than Intel graphics.  If you plan on going with an separate graphics card, you probably won’t see much different.

I built a new desktop a few months ago and when with an Intel quad core processor.  It works great, as I’m able to truly multitask with it.  I wrote a program that did cloud computing.  I was running eight processes at once, and it didn’t slow down my browsing, blogging or email.

wcw/1: if you are asking me what I will do with my current computer, I’m going to give it to my boy, and it’ll also act as a backup.

Andrew, desktop.

***

I need power. Raw naked power.  And memory.  I don’t know how to best balance that cost wise.  If let’s say I was going to spend 1000$, what different configurations on processor and memory would be ideal?

Ars does periodic system guides[1] that cover three ranges: $600-$800, $1200-$1400, and “unlimited”. For each range they offer several alternatives to various parts (often CPU related) if your price point is just outside of any of the ranges. I find them very useful as a starting point.

I would second Andrew/#2. Right now Intel’s i7s are unmatched in performance.

[1]: http://arstechnica.com/gadgets/guides/2010/09/the-ars-system-guide-september-2010-edition.ars

Sean, sweet.  That ARS looks right up my alley…

Are you going to be writing a lot of the code for the Raw Naked Power yourself?  If so, you might consider getting a couple of Nvidia graphics cards and writing code for the Physx engine.

For Ultimate Power I suggest building your own.  I actually just emailed you at the tmapress address a quick and dirty newegg.com list of parts.  It was $1300 but included everything but a mouse and keyboard.  It had 8GB of ram, a 64GB SSD, plus a 2TB HD.  If you aren’t running windows, you can remove that.  If you don’t need a monitor you can remove that too.

Andrew: yes, I write all my code, but I don’t need a graphics card.  I write code to process data, not to create games or other kind of graphics-intensive things.

Can you reforward what you sent to here:
tom~tangotiger~net

Thanks again, much appreciated!

Well the Physx engine on the graphics card allows you to write none graphics programs that are math intensive that run super fast.  For the kinds of things a graphics card is good (vector math, physics, etc) it is extremely fast.

I forwarded the list to you.  If you don’t need a video card at all, the processor has a basic one built in that is fine for decoding video and basic stuff.

I highly recommend getting a solid state drive for your boot drive.  It makes a very noticeable difference in boot times for the OS and load times and responsiveness for applications.  I don’t know if it makes any difference in terms of computing power for database calculations and such, though.

AMD’s Bulldozer and Llano cores are scheduled to launch June 20 and July 4, respectively, according to the rumors:
http://www.tomshardware.com/news/AMD-Bulldozer-AMD-Llamo-APU-DirectX-11,12346.html
Being an AMD employee, I won’t comment on the accuracy of the rumors.

Andrew hit on one of the key ideas. Going with a Solid State Drive (SSD) instead of a platter-based hard drive will make a huge difference, much more than incremental differences in processor speed.

The hard drive on my four-year-old Mac died recently. It’s a 2.16 GHz machine with the maximum of 4 GB of RAM. Everything seemed okay, even running the latest OS. The speed difference from switching to an SSD makes this like a new machine. You can’t appreciate how much time is lost to a machine reading and writing data until you have an SSD. Most tasks that previously took a while probably happen in a third or less of the time they used to take.

And, frankly, nearly every machine today has a decent processor. And most graphics cards are fine for most people’s needs (since most people aren’t doing high-end gaming, running loads of Photoshop filters, or processing video). So be sure to have a good amount of RAM, and then the one remaining hardware aspect which can help is the hard drive. Since I would think that your most intensive work is database related, an SSD will benefit you more than any graphics-related hardware improvements, and probably more than the difference between, say, a 3 GHz processor and a 3.25 GHz processor.

The one odd thing about SSDs nowadays is where they are in the capacity-to-cost spectrum. We’re in the early stages when high capacities are pretty expensive. But we’re also at the point when prices should be dropping relatively quickly. Maybe now you go for a lower capacity and plan to upgrade in two years. Or maybe you spend the money for high capacity now and try to make it work for the four-year life of this next machine.

Likewise, even if you get a platter-based drive, whether as your primary storage or for secondary storage, get one that runs at least at 7,200 RPMs. Again, this makes a huge difference in read/write speeds compared to the common 5,4000 RPM drives. It won’t match an SSD, but the speed gains are worthwhile.

I’m also shopping. (My THT check comes at the end of the month!)

Primarily I do database work, in excess of 50 gb of data in MySQL.

Also, I have a workstation I bought a few years ago to do mapping at home, but the market dried up with the recession and I have had any independent subcontracts in about three years. However, I’d like to keep the ability on the new computer, especially if my office will ever decide to let me telecommute. The main requirement here is a 3d graphics card. The Nvidia Quadro FX 4400 is no longer their top of the line, and has dropped to about $200. The computer I’m on right now in the office has the Quadro 4600, with Intel Core 2 @2.93Ghz, 2 GB of RAM, and gets good performance when only mapping, but if I run db while mapping lack of ram slows it down.

So, I’m looking for quad core, 3+ Ghz processor, 4+ gb of ram, 1+ tb hd.

With those minimums, I’ve found listings on Ebay for $600. I have a Windows 7 64 bit CD at home, and can add the video card later.

Brian/13:
4+ is a low minimum, too low in my opinion. If you’re regularly working with huge databases, it would behoove you to get as much RAM as possible. Like 24 GB worth. It’s an insane amount, but you can get it for ~$300 now.

After I posted #13, I went back to Ebay to check out the i7 prices, appears to add at least $200.

I saw one listing, which linked to the store’s custom build form, where they have the i7, 8 gb ram and 1 tb hd for about $800

I’ve been debating about getting a new machine for myself as well, and I can’t decide if my sabermetic pursuits are sufficient reason to justify getting something ridiculous.  One thing I’ve been toying with is Amazon Web Services...specifically their Elastic Compute Cloud (EC2).  They let you rent virtual machines by the hour, and they have some pretty suped-up options available (for instance 8-core Xeon’s with 64 GB of RAM).  I thought maybe this made more sense since I could pay for what I use (since the majority of the time I don’t need a tremendous machine).  I think they had these in mind more for business and to act as servers, but I can Remote Desktop in and have the power of a server-caliber machine at my fingertips.

I’m still undecided if this makes sense long-term.  For one, I’m pretty sure this is a very atypical use-case of their services, so there is a strong chance I’m not doing things in the way that makes the most sense.  I’m also no DBA, so I doubt I’m getting the most out of MySQL.  One thing I also wonder is how big of an impact an SSD would have on my SQL queries...I don’t have a great feeling for what it needs to do in RAM and what it needs to access the disk for.

Colin I think reported fantastic results with the SSD.  Sometime in late November, I asked people to do some performance tests on the REtro database.  LEt me find that thread for you.

http://www.insidethebook.com/ee/index.php/site/comments/retrosheet_and_computer_performance/

Colin loaded in under 7 minutes (with SSD) and it took me 49 minutes.

Ryan JL commented here, and Colin chimed in two comments later.

http://www.insidethebook.com/ee/index.php/site/comments/retrosheet_and_computer_performance/#32

Are the SSDs only available for laptops?

With a desktop, if I store the db files on a large external USB drive, what kind of improvement would I get over having them on the standard HD platter?

mickeyg/16:
Others have done that. Some because they have a Windows desktop and want a fast Linux desktop/server to use. If you would rarely use the beefier machine, then it does make some sense to use a cloud computer.

***

Brian/20:
No, you can get SSDs for any computer.

Depends on what the two drives are. USB will be significantly slower than an internal drive, but you might be able to get an edge if the USB drive is an SSD and you have a highly random workload. Then the USB drive, by virtue of being an SSD with excellent random read performance, could nudge ahead. But that only applies if it’s a read-dominated workload. Once you start doing a lot of writes, anything internal will be faster.

SSD’s can be put in any computer but you should be aware that they are 2.5”

When I first searched Ebay for ‘SSD’ everything I looked at was for laptops. Then I added desktop to the search and saw that they have kits to add the SSD, many times in a 3.5 floppy drive bay.

I currently have about 50 gb of MySQL files, although some of them are junk tables and can be cleaned out, but if I want to keep the db and OS files on a SSD I figured it would be best to get 128 gb. Looked like I could get a desktop SSD kit for $150-200, but the sellers who offered them as upgrades to the PC being bought were asking $250-300.

I did add an i7, 8 gb ram, 1 tb hd to my watch list ($625)

I mounted my SSD the fancy way.  I taped it into a 3.5-inch drive bay with masking tape.  Since the SSD doesn’t have moving parts, how it’s mounted is not very important.

I did a similar thing as mike.  Sort of screwed it in one side and taped something to it as support.  You don’t really need a “kit.”

The Vertex 2 is a great product that can be had for a little over $200:

http://www.newegg.com/Product/Product.aspx?Item=N82E16820227551

Definitely go for an Intel iX series CPU, Sandy Bridge (2000 series) version.  The i5-2500k is really good, particularly if additional threads aren’t that useful to you.  It’s just about as fast as the i7 and significantly cheaper.

I also recommend an SSD for you, preferably a Sandforce-based model (they’re the best right now).  That Vertex 2 is such a model, but the newer Vertex 3 is significantly faster (though more expensive).  AnandTech Review: http://bit.ly/fkwdkC

Quite a bit of RAM would be good as well.

I really appreciate all the advice guys, here, and via email.

I had not even considered building my own, being happy with whatever custom-built from HP or Dell.  But, seeing that I will need the SSD, it looks like I will have to build my own (my first one!).

Insights on THAT endeavour would be appreciated too.

TT, when something is ‘CPU bound’ that means that the bottleneck is the CPU.  People often think the bottleneck is always the CPU, but they are wrong.  Disk is more often the culprit, which is where a SSD will be a big, big win.  You say, for example, that you want ‘..power.. [a]nd memory,’ which still doesn’t tell us enough.

Take our friend with the 50G mysql db.  [I have my own set of advice for him, but let’s leave that lay.] He is almost certainly disk bound.  Putting a DB partition on a $150 SSD will make all his SQL stuff work much, much better.  Or take some of the stuff I do when I am bored: I tend to load up big data sets and play with them using programs like R.  For this I mostly want gobs and gobs of RAM, because R only works in memory, and a single very fast core, since R does not handle multicore gracefully out of the box—which problem is hardly unique.  But you may be doing embarrassingly parallel work, or running something that scales across cores well.  All these things matter.

Once you figure out what matters to you—single-core power, multi-core power, disk, RAM, GPU, all of the above—it’ll be a lot easier to point you at something.

I will say, Dell and its friends learned a while ago that their profits are higher if they cut costs by reducing component quality.  There was a time I would always have recommended off-the-shelf boxes.  That time has passed, at least in the consumer space.  There I would stick with better quality enthusiast kit, a half-generation behind and deeply on sale.

Some general advice for getting the best performance out of MySQL:

* The number of CPU cores really isn’t that important in a single-user context - running a single query will not benefit from multiple cores. (MySQL will use multiple threads if it’s executing multiple queries at once, but for what Tango and Brian are talking about, I doubt they’re doing much of that).

* RAM. RAM RAM RAM RAM RAM. And remember - you need a 64-bit OS and a 64-bit version of MySQL to take advantage of all that RAM; MySQL is single-process and thus is limited to 2gb of RAM if it’s running in 32 bits.

* Out of the box, MySQL is not going to take advantage of that extra RAM; the default configuration files for MySQL are terrible. On Windows you’re looking for a file called my.ini, in Linux it’s my.cnf. MySQL comes with a file called my-large.ini (or cnf); back up the existing ini/cnf file, copy my-large.ini over it, and restart your MySQL server.

* Partition your databases once they get past a certain size. I cannot stress this enough. You need at least MySQL 5.1 for this.

* The benefits of an SSD have been well covered here, but let me weigh in as well in support of them.

Take our friend with the 50G mysql db.  [I have my own set of advice for him, but let’s leave that lay.] He is almost certainly disk bound.  Putting a DB partition on a $150 SSD will make all his SQL stuff work much, much better.
=============

I just order the Vertex 2 120 gb drive ($195 at Amazon), and will add this manually to the new pc (when it arrives) as the boot drive.

Yes, I have about 50 gb of data in MySQL. Maybe 30-40% of these are temp files, that when IK was doing code development I didn’t want to delete, but they really don’t need to be saved. I would welcome any suggestions in addition to what Colin already offered.

In the mear future, I will continue to run Oliver on the old computer until the new one is fully setup, and then switch over.

I’ve never built my own machine, so I’m not really qualified to judge this overview of how to do it, but the site usually provides good information (nds seank already linked there upthread).

http://arstechnica.com/ask-ars/2011/04/how-to-build-your-own-computer-ask-ars-diy-series-part-i.ars

Fantastic, thank you!

With taxes all done, my time has opened up, so it’s now time to get that new PC.  Thanks again…

Have a problem guys.

Got the new computer, was installing software and data.

Got a blue screen while loaded a large csv file into MySQL 5.1. When I came back, the MySQL would crash whenever I tried to reference that table. Now, MySQL won;t stay running at all. I’ve uninstalled and deleted all files, installed again, and still same thing.

I can start from scratch and re-install Windows 7 in necessary.

I believe I have it fixed.

Even though I uninstalled MySQL 5.1 and deleted the folder, it was still there, perhaps in the registry, so that attempts to reinstall the same version did not eliminate the corrupted installation.

I then tried v5.5, and being a different version it offered installation options that did not appear when I tried reinstalling 5.1. At the end, when the 5.5 installer wanted to start the MySQL server, it refused because one was already running, the corrupt 5.1 that I had tried to remove. I then went into administrative tools and manually disabled the 5.1 service, labeled MySQL, then started a new 5.5 serviced labeled MySQL5. Looks good now, have to go back and import my db’s from sql dump again.

back to Tango’s original question of AMD vs Intel, here’s some CPU benchmarks.

I got the Intel i7 that was recommended in this thread, the 870 running at 2.93 ghz.
http://www.cpubenchmark.net/cpu.php?cpu=Intel+Core+i7+870+@+2.93GHz

I just want to say thanks again to all the Straigh Arrow readers, especially those who made me aware of the SSD.

In the end, I had to go with a laptop.  Considering my home setup, have two desktops would be too much fixed clutter.  The portability of the laptop, both inside my own house (so I can use it and watch TV upstairs), or just if I have to go out, justified it (for me).

Anyway, 160 GB SSD, with 500 GB HD. 8GB RAM.  I-7.  Toshiba, HP, Lenova all allowed for easy configuring, with an SSD (once you knew which of their dozens of models allowed the SSD).

Sorry Mike, but none of them offered AMD as the choice.

Looking forward to getting the machine and loading up RetroDB in November.

I had forgotten about this thread for a while.  As I mentioned earlier, I had been toying with Amazon’s EC2 to rent beefy servers my heavy lifting.  However, that proved to be more expensive than I realized, and I decided that the cost of buying a new machine amortized over a few years was cheaper per month than what I was spending on Amazon.

This past week I ordered a custom desktop from a smaller online builder.  They gave me more flexibility than the likes of Dell but I didn’t need to actually assemble the parts (they do a lot of testing before shipping).

Relevant specs include Intel i7 2600K (overclocked to 4.5-4.8 GHz), 16 GB RAM, 128 GB SSD, and 1TB HD.  It’s still being tested by them before they ship it, but I’m looking forward to playing with it

Feel free to post their website url.

I didn’t want to seem like I was spamming, but I bought from Digital Storm (http://www.digitalstormonline.com/).  Oddly enough, their “gaming” machines seemed more appropriate for my purposes than their “workstations.”

Since I haven’t yet received my machine, I can’t vouch for them, but I did read a lot of good things about them.  One thing I especially liked was that they will overclock and do the requisite testing for you.

Intel Core i5 2500K is a great value...price vs. performance. 

Make sure that if you want more than 3 gigs of RAM that you are installing an operating system that supports it.  Vista and prior will not.  Lower priced versions of 7 will not (e.g., you could buy and install 8 Gb of RAM but if you run Vista, you’ll only get to use about 3 of it).

This site is great:  http://www.cpubenchmark.net/.  Shows speed in relation to price.