AMD 64 question.

[basslover1]

After months of waiting I'm finally getting all the parts in for me new rig. One question arises, my friend mentioned something about the Amd 64 not being able to run at it's optimum unless I use a new version of XP that was made for 64-bit CPU's. Is there any truth to that? Does he know what he's talking about. Because I've got Win2k Pro right now, and don't really want to spend 200+ bucks on a new OS when I've already got this on sitting here doing nothing.

Thanks.

 

[Boonesmi]

even though a64 is a "64bit" cpu, it only runs as a 32bit cpu on most operating systems (the exceptions are the new "winxp 64bit" and some linux builds) so if you use the regular winxp, win2k, win9x, etc then it will be running at 32bit.

the good news is that even as a 32bit cpu its still the best cpu you can buy

the better news is that down the road when microsoft releases its 63bit operating system things only get better


if you want to run at 64bit now, then you need to download the trial/demo version of 64bit winxp (this is not a final version, getting ready to be released sometime next year) or you have to go with linux

 

[imported_michaelpatrick33]

Your friend has no idea what he is talking about. The AMD64 is as fast or faster in 32bit mode vs. the rating equivalent Intel processor. Example: The AMD64 3000+ is as fast and fully compatible with the x86 instruction set (32bit) as the P4 3000 (Northwood or Prescott).

The AMD64 is a complete 32bit processor that runs any X86 32bit OS at full 32bit speed. Think of the AMD64 has both a 32bit processor and a 64bit processor for simplicity. There is no Windows XP64 yet (just a beta) but there are definitely Linux 64bit distros available for your 64bit fun.

If you get Windows XP try and get the Pro version because Microsoft 64bit OS manager (his name escapes me at the moment) stated that Microsoft would in all probability give a free upgrade from XP Pro 32bit to XP64 when it comes out in the first half of 2005

Your Windows 2000 will blaze as fast or faster on the AMD64 as compared to any comparable P4. The AMD64 will definitely run faster than the Athlon XP on Windows 2000 unless you get the AMD64 2800+ and compare it to the AMD XP 3200 then they will run pretty close. (THE PR-ratings are laughable in this regard)

 

[AnotherGuy]

It will work great on windows 2000 which is 32-bit. Your friend is saying that this chip besides 32-bit, is also 64-bit compatable. But the 64-bit Windows XP that is out now is only a BETA ( not a complete version) and i believe this BETA is free to download, not sure. But you can definitly use the one u already have and it will work perfectly... actually very few plp use the 64-bit BETA windows release with these chips.

 

[Boonesmi]

by the way, if you are wanting to try the trial version of winxp 64bit (it is a 1 year trial version) but you can get it here

http://www.microsoft.com/windo...valuation/upgrade.mspx

 

[basslover1]

Cool. You guys are damn quick at responding.

Figured it would run flawlessly, since it runs pretty kickass right now on my Celeron 733 system.

Can't wait to get this thing set up. Probably have it all ready on Friday, and I'm sure I'll be posting questions left and right since this is my first build.

Thanks.

 

[imported_michaelpatrick33]

Originally posted by: basslover1
Cool. You guys are damn quick at responding.

Figured it would run flawlessly, since it runs pretty kickass right now on my Celeron 733 system.

Can't wait to get this thing set up. Probably have it all ready on Friday, and I'm sure I'll be posting questions left and right since this is my first build.

Thanks.

You will be shocked if you are upgrading that far at the difference in performance. What about hardrives? That could slow you down some if you are using a 5400rpm drive.

 

[basslover1]

Originally posted by: michaelpatrick33

Originally posted by: basslover1
Cool. You guys are damn quick at responding.

Figured it would run flawlessly, since it runs pretty kickass right now on my Celeron 733 system.

Can't wait to get this thing set up. Probably have it all ready on Friday, and I'm sure I'll be posting questions left and right since this is my first build.

Thanks.

You will be shocked if you are upgrading that far at the difference in performance. What about hardrives? That could slow you down some if you are using a 5400rpm drive.

7200RPM 200GB Western Digital w/ 8mb buffer. Don't think it'll be a problem.

 

[uOpt]

There is are two articles in the Linux section of Anandtech comparing AMD64s driven with a 64 bit version of Linux versus a comparable 32 bit version.

The 64 bit code improves things quite a bit. But even in 32 bit mode they are not slower than comparable Intel processors for most application, especially games.

 

[WebDude]

I wouldn't say your friend doesn't know what he is talking about. In fact, I think he's right. I've run both windows XP and XP64bit (beta) on my machine, and I get slightly better Sandra scores (there is a Sandra for XP64) running XP64.

However the beauty of the Athlon64 is that it performs outstandingly in both 32 and 64 bit environments. So you can run it now with excellent results, and later down the line when XP64 comes out of beta, you can upgrade to that (if you want). To me it's a win-win situation.

 

[Thermalrock]

ppl keep saying wrong stuff about amds rating system. 3000+ doesnt have anything to do with a pentium 4 3ghz. it means that amd is positive that the 3000+ cpu runs faster than an athlon (thunderbird) would at 3000 mhz if there were a thunderbird 3000 even tho there isnt. last one was a 1400.

 

[Matthias99]

Originally posted by: Thermalrock
ppl keep saying wrong stuff about amds rating system. 3000+ doesnt have anything to do with a pentium 4 3ghz. it means that amd is positive that the 3000+ cpu runs faster than an athlon (thunderbird) would at 3000 mhz if there were a thunderbird 3000 even tho there isnt. last one was a 1400.

The original AMD PR (Performance Rating) system was designed to measure performance of the AMD Athlon processors against Intel's Pentium 4. However, since the P4's performance relative to its clockspeed has changed significantly since then (with much higher FSB and RAM speeds, and new instruction sets, etc.), it no longer matches exactly. Also, you can't directly compare the PRs on the Sempron/Athlon and Athlon64 CPUs.

It certainly doesn't mean that an A64 3000+ runs like a T-Bird Athlon at 3Ghz -- such a chip at even 2.4-2.6Ghz is far faster than *any* AMD '3000+' processor. A 3Ghz Athlon would probably be rated at 3800-4000+.

 

[Thermalrock]

you are totally wrong on that one. i got this straight from the amd site. go check it out. they intruduced it with the core improvements of the athlon xp (palamino) giving him a rating that compared him to the athlon (thunderbird). it was never meant to compare it to a pentium 4. thats what amd says and since theyve invented it ill take their word over yours.

 

[Thermalrock]

ps: btw i cant find it on the amd site anymore but it was up there when they introduced the rating. it would be nice of you if you dont claim stuff thats untrue in future. ima link some inq article for ppl who dont know this.


this is an excerp:

The so-called "PR Ratings" compare the performance of AMD Athlons with Palomino cores against their own processors and are based on 35 different benchmarks, according to a representative at the launch of the chips in Milano today.


and this is the link to the whole thing:

http://www.theinquirer.net/?article=1207


the articles from 2001 and the thunderbird wouldnt have scaled more without some core changes. so while intel was introducing higher and higher frequencies amd started to use the rating to include speed improvements in the new cpus names. this means an athlon xp 1700+ (1466mhz) runs faster than a tbird would at 1700. it DOES NOT mean it runs faster than a P4 1.7. the tbird 1400 outperformed a p4 1.8 already pretty easy.

now stop making crap up and dont dont tell me im wrong when im not.

 

[KoolDrew]

I do not reccomend using Windows 64bit as it is still in beta stages and has many bugs. If you want you could give it a try but it seems like a waste of time to me.

 

[Matthias99]

Originally posted by: Thermalrock
now stop making crap up and dont dont tell me im wrong when im not.

Also from the very article you gave:

While AMD claims that the PR Rating figures are based on a comparison with its own existing microprocessors, there is little doubt that the real purpose of the announcement is to attack the Intel P4 market.

Maybe AMD claims they intended them originally to match up with what the Palomino cores would do if clocked higher (which is pretty goofy, if you ask me), but they've always been compared to Pentium 4 clockspeeds, and it was pretty damn obvious that the only reason they went to a PR scheme was because of the P4's high (relative to the Athlon) clockspeeds.

Edit: Also, even if this was the case, it breaks down pretty badly as the speeds go up. Again, there's no way a 3Ghz Palomino wouldn't completely blow away an AXP 3000+ (which runs at only 2.1Ghz). The T-Bred/Barton cores are faster, but they're not THAT much faster.

 

[Thermalrock]

Originally posted by: Matthias99

Originally posted by: Thermalrock
now stop making crap up and dont dont tell me im wrong when im not.

Also from the very article you gave:

While AMD claims that the PR Rating figures are based on a comparison with its own existing microprocessors, there is little doubt that the real purpose of the announcement is to attack the Intel P4 market.

Maybe AMD claims they intended them originally to match up with what the Palomino cores would do if clocked higher (which is pretty goofy, if you ask me), but they've always been compared to Pentium 4 clockspeeds, and it was pretty damn obvious that the only reason they went to a PR scheme was because of the P4's high (relative to the Athlon) clockspeeds.

the reason they went to the rating system was the high clock of the p4 yes. they wanted to to show that their athlon xps were faster than their tbitds by a higher margin than the mhz would have made ppl think. so they rated it like the TBIRD, NOT PALAMINO would have to be clocked to perform the same.

following your logic i can say its pretty damn obvious youre wrong and they introduced the rating to sell more cpus. the company introduced a rating and let ppl know what it means. the fact that they introduced the rating to sell more cpus in a world of ppl buying high frequency p4s does not change what the rating number of a cpu means.

i made a post you responded to my post saying i was wrong which i wasnt added untrue claims about how thunderbirds at higher theoretical clocks would perform and how amds rating compared amd chips to intel chips. i asked you to not do this anymore and gave you a link cuz i couldnt find the amd statement on their site anymore that was up there when they introduced the rating. you could have just admitted that you saying i were wrong was a mistake but you didnt instead you add valueless information. its natural that a company wants to make money and doesnt want potential customers to think their product is less powerful than the competitions when in fact its not. THE REASON THE RATING WAS INTRODUCED IS NOT WHAT THE RATING MEANS.

i dont feel like arguing anymore just dont tell ppl wrong stuff and dont tell guys they were wrong when you think but dont know they were.

 

[dnoyeb]

Thermalrock you are correct, now let it go.

I personally will not be buying any AMD 64bit chips as I can't see the need. What is going to push this chip? I already played doom3 on my XP3500+ or whatever I got (Barton Core 400mhz bus). See I can't even remember cause they all so fast these days. Im running 9700pro ATI.

The bump is not big enough for me to come on board yet. I'll probably buy in July '05.

 

[alexruiz]

Keep your windows 2000. The K8 loves it! I reccomend a dual boot partition with your favorite 32 bits OS (win2K is my vote) and the beta of the XP 64 bits.

Win2K proved to be faster, more stable, more flexible and less intrusive than XP (both home and Pro). I used them both in my M6805, and win2K was consistently faster with the exception of boot time. Make sure to get a tweaking utility to make win2K see the full amount of cache (Fresh UI for example) Gaming performance is equal between XP-2K, office performance is better in 2K, multimedia performance is better in 2K (make sure codecs and settings are well tuned). 2K forgets less settings, and isn't restricted about file search options.

As wild card, the 64 bits edition is faster, and not by little in video encoding. It can go from 6-7% in TMPGEnc to almost 30% in ulead video studio (version 5, I have to test it with version 8). Nothing different, just a different OS. And if you want your machine to turn imprssive snappiness and almost instant boot, try winME. It feels faster than anything, and bot time is around 13 seconds (from power on button push to desktop and no hard drive activity)

Win2K + XP64!


Alex

 

[Matthias99]

Originally posted by: Thermalrock

Originally posted by: Matthias99

Originally posted by: Thermalrock
now stop making crap up and dont dont tell me im wrong when im not.

Also from the very article you gave:

While AMD claims that the PR Rating figures are based on a comparison with its own existing microprocessors, there is little doubt that the real purpose of the announcement is to attack the Intel P4 market.

Maybe AMD claims they intended them originally to match up with what the Palomino cores would do if clocked higher (which is pretty goofy, if you ask me), but they've always been compared to Pentium 4 clockspeeds, and it was pretty damn obvious that the only reason they went to a PR scheme was because of the P4's high (relative to the Athlon) clockspeeds.

the reason they went to the rating system was the high clock of the p4 yes. they wanted to to show that their athlon xps were faster than their tbitds by a higher margin than the mhz would have made ppl think. so they rated it like the TBIRD, NOT PALAMINO would have to be clocked to perform the same.

following your logic i can say its pretty damn obvious youre wrong and they introduced the rating to sell more cpus. the company introduced a rating and let ppl know what it means. the fact that they introduced the rating to sell more cpus in a world of ppl buying high frequency p4s does not change what the rating number of a cpu means.

i made a post you responded to my post saying i was wrong which i wasnt added untrue claims about how thunderbirds at higher theoretical clocks would perform and how amds rating compared amd chips to intel chips. i asked you to not do this anymore and gave you a link cuz i couldnt find the amd statement on their site anymore that was up there when they introduced the rating. you could have just admitted that you saying i were wrong was a mistake but you didnt instead you add valueless information. its natural that a company wants to make money and doesnt want potential customers to think their product is less powerful than the competitions when in fact its not. THE REASON THE RATING WAS INTRODUCED IS NOT WHAT THE RATING MEANS.

i dont feel like arguing anymore just dont tell ppl wrong stuff and dont tell guys they were wrong when you think but dont know they were.

1) Run-on sentences with no capitals are *really* hard to read.

2) You still haven't addressed the fact that the ratings make no sense at ~3000+ speeds -- a 3Ghz Tbird (sorry I got the two chips backwards in my description; I knew what I was thinking about ) would blow away a 2.1Ghz Tbred/Barton -- they're just not that much faster. The ratings make no sense unless you apply them against the P4s, at least once you get above 2000+ or so. Even if AMD originally intended the ratings to compare the AXPs against the older T-Bird Athlons, the ratings are largely useless for doing that, and the ratings match up a lot better to P4 speeds than they would to hypothetical T-Bird speeds. AMD's 'official' description of the rating is not necessarily what it really means.

3) Chill out, man. Have a drink or something; you seem tense. :beer:

 

[Thermalrock]

why would you make a claim that a 2ghz barton would get blown away by a tbird at 2ghz. i already have too many drinks the way it is doesnt mean i will skik this one.

 

[Matthias99]

Originally posted by: Thermalrock
why would you make a claim that a 2ghz barton would get blown away by a tbird at 2ghz. i already have too many drinks the way it is doesnt mean i will skik this one.

Read my post again, more carefully (maybe *less* drinks are the answer here ).

You're claiming that the PR is supposed to mean "An AXP processor with a PR of X+ runs at the same speed as an Athlon T-Bird running at X Mhz". If this is true, an AXP Barton 3000+ (which runs at 2.1Ghz) should be the same speed as a T-Bird running at 3Ghz (ie, 3000Mhz). Somehow I find this hard to believe -- I'm almost certain you'd need at least 2.5-2.6Ghz out of the Barton to match it -- but that same Barton 3000+ is just about as fast as a P4"A" running at 3Ghz.

 

[Thermalrock]

youre wrong. the northwood there is no willamette running at 3ghz isnt slower than a tbird would be at 3ghz. and you dont need a barton to run at 2600 to be as fast as a tbird would be at 3ghz. if you believe it or not doesnt matter.

 

[Matthias99]

Originally posted by: Thermalrock
youre wrong. the northwood there is no willamette running at 3ghz isnt slower than a tbird would be at 3ghz. and you dont need a barton to run at 2600 to be as fast as a tbird would be at 3ghz. if you believe it or not doesnt matter.

These are the speeds and ratings of the AXP Tbred and Barton processors, and how much faster they are rated than their actual clockspeed:

166Mhz T-Breds (256KB L2, 133Mhz FSB)
AXP Tbred 1.33 -- '1500+' -- 12.5% faster
AXP Tbred 1.40 -- '1600+' -- 14.3% faster
AXP Tbred 1.47 -- '1700+' -- 15.6% faster
AXP Tbred 1.53 -- '1800+' -- 17.6% faster
AXP Tbred 1.60 -- '1900+' -- 18.8% faster
AXP Tbred 1.67 -- '2000+' -- 20.0% faster
AXP Tbred 1.73 -- '2100+' -- 21.2% faster
AXP Tbred 1.80 -- '2200+' -- 22.2% faster
AXP Tbred 2.00 -- '2400+' -- 20.0% faster
AXP Tbred 2.13 -- '2600+' -- 21.9% faster

333Mhz T-Breds (256KB L2, 166Mhz FSB):
AXP Tbred 2.08 -- '2600+' -- 25.0% faster
AXP Tbred 2.17 -- '2700+' -- 24.4% faster
AXP Tbred 2.25 -- '2800+' -- 24.4% faster

333Mhz Bartons (512KB L2, 166Mhz FSB):
AXP Barton 1.83 -- '2500+' -- 36.4% faster
AXP Barton 1.90 -- '2600+' -- 36.8% faster
AXP Barton 2.08 -- '2800+' -- 34.6% faster
AXP Barton 2.16 -- '3000+' -- 38.9% faster

400Mhz Bartons (512KB L2, 200Mhz FSB):
AXP Barton 2.10 -- '3000+' -- 42.8% faster
AXP Barton 2.20 -- '3200+' -- 45.4% faster

Now, the ratings on the TBreds don't look *so* unrealistic, although the difference between the 2000+ and 1500+ is, you have to admit, suspicious. Somehow a 25% increase in clock speed (with no other changes) led to a 33% increase in rated speed. Either they artificially lowered the ratings of the low-end processors, or they're inflating the ratings of the higher-end ones. In either case, it doesn't look like they were real concerned about making sure the ratings matched up with what a T-Bird would do.

I'd love to find benchmarks that compare the number-crunching power of a Barton with a 166Mhz bus heads up against a T-Bird with a 133Mhz bus, but I'm having difficulty finding any right now. I just have trouble believing that the Barton processor is really almost 40% faster clock for clock than the T-Bird, which is what would be required to make those ratings accurate. It also brings up issues of what exactly they measured -- integer performance? Floating point performance? Both? Application performance (which is dependent on chipset and memory as well...)?

Also, here's a link from AMD where they directly compare the 3000+ and 3200+ against the performance of the P4 -- nary a peep about the old Athlon T-Bird. comparison

Here's something from SetiAtHome link -- a T-Bird 1.46Ghz versus a TBred 1.67 (same FSB) shows only a 12% increase in MFlops/sec. when scaled by clockspeed (that is, if you assume a T-Bird would scale linearly in this regard with a clockspeed increase).

Some POV-RAY benches: a T-Bird 1.40 versus a TBred 1.40 (AXP 1600+) shows almost no improvement -- it's less than 1% faster. A Barton 2500+ at 1.83Ghz (on a 333Mhz FSB this time), scaled by clockspeed differences, is, again, only about 1% faster.

Another review where they put the Tbred and Barton head-to-head at the same clocks. The Barton performs identically in synthetic tests, but is a few percent faster at the same clocks in gaming tests (possibly because of better OOE, and/or the larger cache). They did at least reference that the PR is *supposed* to be related to the speed of an Athlon T-Bird.

I would have to agree that AMD originally intended the PR to correspond to 'effective T-Bird speed', although how to measure that accurately seems unclear (since most simple synthetic benchmarks show almost no difference clock-for-clock). You were right about that. However, there is little evidence that they have necessarily stuck with that, and their ratings seem to have gotten considerably inflated with the faster AXP processors.

Of course, I don't know what any of this proves.