Atomic I/O letters column #115
Originally published 2010, in Atomic: Maximum Power ComputingReprinted here April 11, 2011 Last modified 16-Jan-2015.
Do 47 Rage 128s equal one HD5970?
In Atomic whenever a motherboard is reviewed, if the second PCIe slot is limited to x4 it is suggested that you don't use it for CrossFire. But 8x is OK?
I am wondering what the actual CrossFire speed difference is between x4, x8 and x16 in the second PCIe slot?
Henry
Answer:
This is a bit of a how-long-is-a-piece-of-string question, because different people have different opinions about
what "OK" performance is. To a large extent, it depends on what you're using CrossFire for.
If you're upgrading an older ATI-graphics-card computer by adding another card, you should see a respectable performance gain even if you end up using one x16 and one x4 slot. (Provided, of course, that your system was really actually limited by the speed of the graphics card in the first place.) CrossFireX, like Nvidia's SLI, has a separate hardware data link between the cards, so they don't have to push their brain-synchronising data through the PCIe bus, and a mere four-lane PCIe slot is thus less restrictive than you might think.
(Before CrossFireX there was a more limited hardware-linked version, and then "software CrossFire", which had no separate bridge connector between the cards, and so did push all of the inter-card data through the system bus.)
Even with CrossFireX's ability to combine different cards in the same model series (so, any 48xx card with any other 48xx, and any 38xx with any other 38xx...), though, the economics of this can be questionable compared with just saving up to buy one faster card in the quite near future. But plenty of people have given good homes to two-year-old Radeons from which someone else upgraded. Going from one 3850 to a pair of them will make a difference, even if one is in an x4 slot.
Just don't spend big bucks on a new card that's going to end up in an x4 slot; that was a lousy idea in 2008, and hasn't improved since.
Your traditional CrossFire/SLI user is someone who wants really high frame rates at really high resolutions with really high graphics-quality settings on really recent games. Remove even one of those requirements - use a lower resolution, or wait a year before upgrading your graphics card and playing a given game - and you slide a long way back in the financially-sensible direction on the price/performance graph, and often won't need a multi-card system at all.
But if you want it all and you want it now, putting your second CrossFire card in an x4 slot will cause you to waste even more money than you intended to.
The x16-versus-x8 difference, for a second CrossFire card, is less pronounced. This is in accordance with the ancient dictum that each single increase in graphics-card-slot bandwidth - ever since AGP version 1 gave us the first "modern" graphics-card slot - hasn't made much difference to anything. The difference will be readily measurable (at high enough resolutions, quality settings, et cetera), but it takes a big bus-speed difference to have a significant impact on graphics performance. Unless you've run out of RAM on the graphics card and data's being thrashed to and from main memory, that is, in which case faster bus speeds will turn an extremely lousy frame rate into a merely very lousy one.
Buck the trend: Downgrade to Win2000!
Is there now any reason not to switch from Windows XP 32 bit to Windows 7?
Am improving cooling by leaving the case open with HEPA filter blowing at it (I have asthma). Upgrading my Dell Inspiron 530 to a Corsair 550W PSU, and my Core 2 Quad Q6600 to a Q9650. Graphics card is an HD 4550; will add memory as needed.
Robert
That "Try Again" button may be worth the Win7 purchase price all by itself.
Answer:
There aren't any really good reasons not to upgrade. Win7 generally runs well even on older hardware, and there
are some real improvements.
This isn't a guarantee that your upgraded computer won't have some problem that it currently lacks, but it can still be refreshing to visit the Problem Exchange and trade your old boring WinXP problems for exciting new Win7 ones.
I, in general, don't recommend upgrading anything if you don't have a real reason for doing so. On the face of it this is pointless advice like "omit needless words", but what I mean by it is that people often get the new OS just because it's the new OS, and routinely upgrade a graphics card when what they need is a faster CPU, or upgrade their whole system when all they needed was a new graphics card, et cetera. A lot of people would be much better served by leaving their PC as it is and buying a bigger monitor than by upgrading other stuff. I've had my 30-inch Dell for more than three years now, and it's definitely the best money I've ever spent on computer gear.
If your system is for one reason or another in need of an OS reinstall anyway (even if you're only upgrading your worryingly-old boot drive and can't be bothered imaging the old one onto the new), then upgrading to Win7 is a perfectly sensible thing to do. But the only real class of software that you can't run on WinXP (without HEINOUS kludges...) is DirectX-10-and-higher games.
DX10 was a real, important step forward, not because it allowed games to be spectacularly prettier but because it made a clean break from the terrifying tangle of workarounds and special cases that had seeped into Direct3D over the years. This has created a system that's much easier to write widely-compatible software for - which means more, and better, games. The price we pay is the XP-to-Vista break, Win7 being essentially Windows Vista 1.5 We're Sorry About 1.0 But You're Still Not Getting A Refund Edition.
I also make absolutely no representations about what hardware-upgrade pitfalls may be lurking in your Dell - but, these days, you'll probably be fine. Dell no longer pack (most of) their computers with wilfully-nonstandard hardware, so you can indeed do crazy far-out stuff like upgrading your own PSU and CPU with off-the-shelf items. I still never so much as touch a screwdriver to a Dell without downloading the service manual, though. For all I know, they're just trying to lull us.
In the specific case of upgrading the CPU of a non-overclockable brand-name box, bear in mind that putting your current, famously-overclockable Core 2 Quad CPU onto a regular overclockable motherboard can get you the same result, and then some. A Q9650 shouldn't cost you a great deal at the moment, of course, but if you're only running at stock speed then the 1.25X CPU performance difference over a 6600 won't necessarily have much visible effect.
(I became an expert on this at quite a young age, when I upgraded my Amiga 500 from a 68000 to a 68010. Some of you are laughing now.)
Again, I'd definitely put the money toward a big-ass monitor first!
Burning questions (no, I couldn't think of a better title)
I've just bought a second-hand car with a pretty good but not-replaceable sound system. It's a few years old (built in 2002) and only plays audio CDs, but I figured "well, I'll just burn some CDs from my perfectly legitimate collection of music on my PC, and that'll be good enough". Problem is, it's having trouble reading burned discs - of the five discs I burned to try it out, only one would ever be picked up properly and start playing, and it only ever got through the first one and a half tracks before getting confused and skipping.
I suspect it's just the head unit getting a bit old, and I need to get the CD reader cleaned or replaced, but looking around on the net for people with the same car and problem I've seen some advice along the lines of "burn your discs at the slowest speed you can, that should help make them more readable". This struck me as utter crap in this day and age; shouldn't all modern PCs be able to burn CDs as fast as they can without having problems? Practically every burner sold today has some kind of SuperBurnBufferShieldThing that theoretically eliminates that sort of running-out-of-data-midburn scenario, and I can't see how it'd make the data any more "burned better" to the disc to make a difference; surely if the burn was bad Nero would error out and tell me to try again?
To me, this sort of thing should be a thing of the past, same with DVD plus or minus; customers at my shop often ask for one or the other, because they've been told by a friend that it's the better way, but modern DVD burners work just fine with both kinds, and I've never had anybody come back and tell me they didn't work and that I sold them the wrong kind.
Tim
Answer:
Frankly, I consider it a miracle that burned CDs and DVDs ever work. The reliability that modern drives manage
to get out of an immense series of barely-distinguishable fuzzy splodges is remarkable.
CD-Rs and DVD-Rs are a pretty mature technology today, and as with the later years of the audio cassette, they're doing stuff they were never originally expected to manage. Modern CD/DVD burners are way faster than was initially thought possible; CD-R only topped out at 52X because of vibration, and discs physically disintegrating. With so many different manufacturers of different discs (some of which are different because of actual technological improvements, but most of which are different just to make them cheaper, and/or avoid stepping on some other company's patents), there's a very wide variety of possible burned-disc characteristics. It's unsurprising that some readers, especially older ones, can't read all discs.
The solution, fortunately, does remain the same - try different discs, try different burners (a ridiculous idea when a CD writer cost more than the rest of the PC, but easy now), and, yes, try different speeds too. Burning speed should make little to no difference for data discs, but the error-correction-less bit-stream nature of audio-CD data makes it eminently possible for a tiny change to make a marginal reader completely fail. It's not even very hard to make a disc that starts playing just fine in the car, or some other marginal player, but craps out after a certain number of minutes of music, as the player reads tracks further and further from the centre of the disc.
Note that the various buffer-underrun-preventing technologies all do actually stop and resume the burning process, which was impossible with older writers. They can't perfectly join the resumed write session to the end of the other one, though, and this too can cause problems with some readers. This could easily be a complete disaster for an audio CD, but modern computers are unlikely to run out of data when burning a CD, especially at reduced speed. (I don't actually know which, if any, of the pause-and-resume technologies even work for audio-CD burning.)
Also, remember that you can now get MP3-playing FM-broadcasting doodads. They typically have a card slot and a USB socket, and thus give you access to huge amounts of music without fiddling with multiple audio CDs. Audio quality is usually more than good enough for stock-car-stereo purposes. They're so cheap that you really might as well give one a go. (Don't ask me which one's best, though!)
