The new forums will be named Coin Return (based on the most recent vote)! You can check on the status and timeline of the transition to the new forums here.
The Guiding Principles and New Rules document is now in effect.
Getting a 64bit CPU - Have some questions.
Dublo7
Registered User regular
Registered User regular
Hey guys,
So in about a week or so, I'm getting a new PC, and I have a few questions about the 64bit CPU I'm getting.
So in about a week or so, I'm getting a new PC, and I have a few questions about the 64bit CPU I'm getting.
- Will I be able to run Windows XP Pro SP2 fine? Do I need a 64bit version?
- What's the benefit of installing 64bit applications? Are they just tailored to run faster with the 64bit CPU?
- Is Gigahertz speed still a huge factor when it comes to buying a 64bit CPU? Can you have a lower Gigahertz, but still run things reasonably well? For example; would the difference between a 4000+ AMD AM2 be huge from say a 6000+ AMD AM2?
[SIGPIC][/SIGPIC]
Dublo7 on
0
Posts
As Spoit said, the primary advantage right now is access to RAM above about 3.2GB. In theory, the x86-64 chips running all 64-bit software can have some increased performance as well, mostly due to an increased number of CPU registers as well as an improved instruction set. However, it seems the compiler vendors have not caught up quite yet and so these gains have yet to be fully realized.
The Ghz factor is getting weirder than ever, even in processors from the same company. My Pentium M 1.4 Ghz laptop runs noticeably faster than my old 1.8 Ghz Pentium 4, and my 1.6 Ghz Core 2 Duo at work runs even better than that (well, it does have another core).
Ah, you see, that's the kind of thing I was worried about. Thanks for that.
And yes, the whole Ghz thing is very strange.
A Core 2 Duo will do you more good than that 6000+. AMD can still make a decent rig but Core 2 is where its at for the moment. Cheap, powerful, cool, very easily overclocked.
Shogun Streams Vidya
Windows XP x64 came out in 2005. There were lots of missing drivers and incompatible programs back then. I had to wait months for my TV tuner to get drivers and for a virus scanner that would run in x64. Now a days, the vast majority of hardware and software has 64bit driver support. The whole incompatibility issue that x64 versions supposedly have is way overblown. I can't think of any hardware that doesn't run in x64 (I'm sure there's obscure stuff out there, but I personally don't know of any) and, except for the Cisco VPN just mentioned, I can't think of any programs that will run in 32-bit but not x64 either.
Personal experience, YMMV, yadda yadda, but do keep in mind that a year is a huge amount of time in the computer world. (in 2005, x64 didn't even have virus scanners that worked. 2 years later, almost everything works) If you have something that isn't supported now, chances are it will be very soon. I would go with x64 because while there may not be much benefit today, 32-bit has reached its peak. x64 still has a lot of potential for improvement so it is a better bet for a system that you will likely be using for the next few years.
Also, Intel's new revision of the Core 2 architecture (code name Penryn) is being rolled out. I haven't been paying too much attention to it, but the server grade Xeons are already out, and I believe the desktop processors are being rolled out this month (although it may be in January). These a rather more efficient version of the architecture, with a larger cache, higher bus bandwidth, and a brand new SSE4 instruction set, which makes encoding and decoding video and audio files close to 100% faster. Seriously, some of the bench marks just blew my mind. If you want to pay $$$ for the new revision, or if you do lots of stuff that Penryn excels at (you need low-power, low-heat, or you do a lot of encoding, or whatever), you may want to wait and go with that.
Otherwise, the previous generation of Core 2 processors still absolutely destroy AMD's offerings. I'm an AMD fanboy from way back, and I've been very disappointed. Their new processor, code named Barcelona is pretty much terrible. It has a hardware defect that makes the system unstable when under high load, and it only has between 1%-10% advantage over the previous generation (that is, not Penryn), clock for clock. While the on-die memory controller of the AMD chips is really nice, and does speed things up, you'll need a good chipset and the proper RAM (with good timings) to take advantage of it. As it stands right now, you could get a previous generation Intel processor (like the one I have) at a higher clock rate than the current AMD "Phenom" processors that would be both cheaper and perform better.
HOWEVER, if you currently have a AM2+ motherboard, I am fairly certain that the new Barcelona chips are drop-in compatible.
Urhm, what? It's not like double precision (64-bit floating point) variables haven't been available for a long time...
Aye, the advantage in 64-bit for your common user lies in the 64-bit registers. More register space, means less frequent stack use, means faster computation, means user sees a performance increase. The issue now is that almost all x64 programs are programs written for 32-bit registers, but compiled so that they work on 64-bit registers. So while they run on 64-bit hardware, they don't take full advantage of 64-bit hardware. That should hopefully change and improve with time.
Ganz falsch. First, TechBoy is arguing that somehow 64-bit applications have increased mathematical precision (not performance) over their 32-bit counterparts. This might be true in a world where scientific coders just let the precision of their applications depend on whatever the word width of the machine is. Problem is, nobody does this, since precision is like, important in these applications.
People writing scientific and mathematical applications know whether their variables are 32-bit floats or 64-bit doubles, and a recompilation shouldn't change this. Now, if they decide to recode their application to use 64-bit values internally because they reason that the processors can now handle 64-bit values more quickly, you may see an increase in precision, but only if they go to the trouble of porting the app.
Additionally, you're seemingly arguing that increased performance in a 64-bit machine comes from having wider registers? For a small subset of operations this is true, or if your application does a lot of work on very wide values. This is not where x86-64 gets its performance gains, though. There are three primary reasons why you get increased performance here:
1) They didn't just double the width of the registers, they also doubled the number of registers. These are orthogonal design decisions - there's no reason they couldn't have come up with a 32-bit variant of x86 that had 128 general purpose registers. It's increasing register count, not register width, that lets you keep more variables in the processor rather than in cache or RAM.
2) There are new opcodes that do the same amount of work in a single operation that older instruction sets would do in multiple instructions.
3) Ability to address RAM in excess of 4GB.