2.4 vs 2500

[Jan Kivar]

And finally... are these canterwood motherboards worth the money.. or should I just settle with a springdale. Any observations would be welcome.

Most manufacturers have enabled the PAT in Springdale, which is the only difference between Canterwood and Springdale (the Canterwood supports ECC, but there is very limited amount of 200MHz memory sticks w/ ECC).

Some boards can use PAT only with two sticks, not four. But, buy two 512 MB sticks, and there's no need to upgrade in few years, and then we are most likely using DDR-II sticks.

Cheers,

Jan

[ez2remember]

From what I read and from personal experience, it contradicts with your findings. AMD XP chips performs better than the equivalent Intel P4's in real world tests (MS office apps, photo editing, DTP, web browsing & creation etc.) This is likely what we would be using our comps for. I don't trust many benchmarks because it does not reflect the way we use our computers. From my POV real world tests are the best indication of performance and not just a sheer number indicated by many benchmarks.

e.g. XP 2200+ is faster than the equivalent P4 2.2Ghz. Where the P4 does perform slighlty better than Atlon XP equivalants are in 3D tests.

Even the new 800FSB variant of Intel, much hyped about, but the XP3200+ with 400FSB outperforms the equivalent speed from the reviews I have read.


P4 chips also dissipate slightly more heat than AMD equivalent. The new thoroughbred b and barton cores runs cooler than p4 chips. I have nothing against Intel, I use one myself. It has its advantages and disadvantages but for perfomance and heat, it does not outperform the AMD equivalent.

[jamoore9]

Um, I have to disagree with ez2remember. Nothing personal. If you're going to: game (as I guess you will by your handle), rip or burn CDs, do anything with video, or use web browsing features beyond mere HTML (Java, Microsoft VM, ActiveX), then you want the P4 2.4C. I think he might be comparing the XP2500 to the P4 2.4B. The 2.4C with hyperthreading enabled in WindowsXP is faster than the P4 2.66B, and in some cases the 2.8B.

No, synthetic benches aren't all that indicative. And P4 "B" chips are often outtperformed by AMD equivalents in tests. But the P4 "C" chips do indeed rock when paired with Canterwood and Springdale. As for the XP3200+ vs. the P4 3.2C, check out the brand new comparison at Tom's Hardware: http://www6.tomshardware.com/cpu/20030623/index.html

The 3.2C WALLOPS the XP3200+. Tom's also gives the skinny on comparing the P4 C to the P4 B and AMD equivalents. Clearly no contest: the XP2500 doesn't beat the 2.4C in a single test, not even "real world" tests like PCMark and SysMark.

As far as Springdale vs Canterwood: the ONLY reasons you would want Canterwood are if you can't live without CSA (since not all Springs have it) or ECC. PAT is no longer an issue, since Intel left a "backdoor" to activating it that Asus let out of the bag and now most manufacturer's are exploiting via BIOS updates.

The Asus P4P800 is currently the "king-of-the-hill" in terms of Springdale performance. In fact, in a lot of cases, it the king of ALL performance, since it is faster than Canterwood boards from MOST manufacturers. And its cheap. (The P4P800 and 2.4C combo is what I'm using right now. I can send you some bench results if you want to see what it is capable of -- PM for those.)

ez2remember is right about heat, BTW. The P4C is hotter than an AMD. But keeping your SLK-800, in my view, would be a dumb reason to go with an AMD. The stock Intel cooler is pretty efficient (see MikeC's own review for that) and not terribly loud, so it works well as an interim. Personally, I'm using a Zalman CNPS6500-AlCu with the fan at about 8V, 2.4 OC'd to 2.65: 41C idle, 46C under load.

Hope that helps you clear the mud!

jamoore9

[Mr_Smartepants]

Well I have a P4B 2.4 overclocked to 2.7GHz (DDR400) using a 845PE chipset board. I just finished building a Barton 2500+ nForce2 combo for my dad (see here).

I've just finished the benchmarks and I can say that without a doubt, I'd build this setup for myself next time! The Barton will go toe-to-toe with my P4 no problem! They are so close in performance! They edge each other out in every other benchmark.
Oh, and did I mention the Barton was about $100 cheaper!
Oh, and the Barton runs about 35*F cooler than the P4! Both running an Alpha PAL(8045 vs 8942) with Panaflo L1A @12V

Until the Intel prices start coming down, go with the Barton/nForce2 combo. It's a true bargain!

[miker]

In addition, a Barton will seriously take the wood to a P4 when running Folding@Home. Those FPU's.

[Mr_Smartepants]

In addition, a Barton will seriously take the wood to a P4 when running Folding@Home. Those FPU's.

Heh, heh! He said 'wood'. Heh, Heh!

Sorry, couldn't help myself!

[Mr_Smartepants]

However, 20C of heat is a big difference - are you sure that's right?! I mean, I run at 50C now, implying that the p4 would run at 70C.

Well I can tell you from personal experience that MY P4 runs under 100% load all the time (folding) and the core CPU temp (measured with MBM from the onboard diode) reads 140*F (60*C) and the case temp is 113*F (45*C). The new Barton 2500+/nForce2 combo listed above ran CPUBurn for 6 hours straight and the core CPU temp (measured with MBM from the onboard diode) reads 110*F (43*C) and the case temp is 75*F (24*C). The case temps were verified with a glass/mercury thermometer (stuck in the exhaust fan screw hole with bulb sticking 'In' toward the case-center).
I'm telling you, it's THAT cool!
I couldn't believe it myself until I proved it with the thermometer and 'hand-over-exhaust' method.

[Mr_Smartepants]

Ah, your ambients are so low. My ambients run in the mid 40s.

Still, if you account for the differences in ambiets, the intel runs +15, the AMD runs +19. I think your AMD case has much better airflow.

Yeah, my room temp hovers between 65 & 69*F (18-20*C). On the AMD box, I switched the front fan from the JMC to the Globe fan that came with the case (at 5V on Fanmate). I'm going to mod the Stealth, Globe, and JMC fans with resistors to drop to 5V and retest. That would bring it back in line with my Sonata.

Back to the original post:
If cost is a concern, go Barton/nForce2.
If you've got money to burn, go Intel P4C /865PE Springdale. With Intel, hardware longevity is assured (Socket 478 will stay for a while).

[Wrah]

And I don't think the barton is a better folder then the P4. Look around in the benchmarks at TomsHardware.. He tests Seti somewhere and the P4 are all together in the top half. Now I did read somewhere that Seti is much smaller and relies less on cache then folding, so may not be a good test. Next closest is I think are the benchmarks for video- and audio compression, and winrar. And there too do the p4's leave the athlons in the dust. I can't find anything in the benchmarks which shows the athlons might be faster then the p4's.
If I would have done my homework a little better 2 weeks ago, I might have bought a p4 instead of a barton.

Another pro of the athlons which hasn't been mentioned yet.. it's better imo for overclocking. Because you can change the multiplier you can max out both the mobo and the cpu. If you're interested in that. Or offcourse the oppositie.. let the cpu run cool on a low multiplier but the board on it's normal speed, that's not possible with a p4.

Can also go for the cheap solution, a barton 2500+. And spend the money you save on not buying intel on a cheap matx mobo + xp2100 and let it fold 24/7. Those 2 would be enough to get you around 900-1000 points a week I think.

[Mark Larson]

I don't know how accurate temp measurings are, but:

http://www.sandpile.org/impl/k7.htm
http://www.sandpile.org/impl/p4.htm

[dago]

With Intel, hardware longevity is assured (Socket 478 will stay for a while).

ROTFL

For AMD, socket 7, Slot A (for a very short time) , Socket A (for a > 2GHz range) and soon Socket 940 for the new 64bits.


For Intel, in meantime, you've got Socket 7, Slot 1, Slot 2, FC-PGA (aka Socket 370), FC-PGA2, mFCPGA, Socket 418, Socket 478, Socket 603, mPGA700, Slot M and I probably forgot some or mix some.

[tragus]

With Intel, hardware longevity is assured (Socket 478 will stay for a while).

It's not just CPU-related news that we have to report: both manufacturers are working on new platforms as well, with AMD using Socket 754, and Intel, Socket 775 (Socket T, LGA construction) [[that will use DDR2]] starting Q2 2004. Prior to that point, the Pentium 5 will make its debut on the "old" Socket 478.

So, "a while" is "about a year" right now.

[dago]

Yep, but in any case, Intel tendency to switch socket is >>> than AMD's.

[jamoore9]

With Intel, hardware longevity is assured (Socket 478 will stay for a while).

ROTFL :D

For AMD, socket 7, Slot A (for a very short time) , Socket A (for a > 2GHz range) and soon Socket 940 for the new 64bits.


For Intel, in meantime, you've got Socket 7, Slot 1, Slot 2, FC-PGA (aka Socket 370), FC-PGA2, mFCPGA, Socket 418, Socket 478, Socket 603, mPGA700, Slot M and I probably forgot some or mix some.

I think what he means is that with a Springdale chipset, if you get tired of a P4 2.4C you will have the option to upgrade to the first generation of Prescott processors up to 3.6Ghz and FSB1000 (some Springs are even doing FSB1200 pretty righteously with MAD cooling). That's a lot more headroom than Socket A will give you (max is the 3200+).

Seriously, this doesn't even bare thinking about. Most Springdale boards cost no more than an nForce2 board, anyway. So get a spring and a 2.4C and experience the sweetspot! I'm there, its beautiful!

[Rusty075]

And I don't think the barton is a better folder then the P4.

AMD's thrash Intels when it comes to folding, if that's important to you.
Here's a direct comparison: CPU Speed and Points Information

[Gandalf]

My XP1700+ *trashes* the intell 1.8gig at folding ...

[Wrah]

Hmm.. Then I wonder in what way folding would be so different then from the number-crunching programs they use in the benchmarks. Perhaps the benchmarks are calculating mostly in integers while the folding works with floating point numbers.

And the the projects on that page aren't on stanfords project list, so that page might not be exactly up to date anymore. I wonder if there's even a gromacs project on that list.

[Rusty075]

It's all about the Floating Points. The AMD has 3 fully pipelined FPU's while the P4 has 2 (one of which is dedicated to register transfer)

According to the benchmarks people have reported the Gromac's core is even more floating point intensive, giving the XP's even more of an advantage there. The situation may be about the change however, the rumored new folding@home core is supposed to be SSE2 optimized, which should even things up.

[Jan Kivar]

For Intel, in meantime, you've got Socket 7, Slot 1, Slot 2, FC-PGA (aka Socket 370), FC-PGA2, mFCPGA, Socket 418, Socket 478, Socket 603, mPGA700, Slot M and I probably forgot some or mix some.

I'm sure You meant Socket 423, didn't You?

Jan

[Beyonder]

Um, I have to disagree with ez2remember. Nothing personal. If you're going to: game (as I guess you will by your handle), rip or burn CDs, do anything with video, or use web browsing features beyond mere HTML (Java, Microsoft VM, ActiveX), then you want the P4 2.4C.

I do all of that with an old XP1500+....and to be perfectly honest, I was doing everything except gaming with an old PII-400.


I think the best advice to to buy the processor at the best pricepoint, irregardless of whether or not it's Intel or AMD. If gaming is your emphasis, then everything I've seen puts the AMD's and Intel's pretty equivalent, and more critical is the video card--so I'd go with the most bang for the buck.

I checked: I'd get either the Barton 2800 or P4 2.4C...I'm guessing overall the barton is probably a bit faster, but to be honest I'd doubt anyone here could tell a real world difference between the two. Both of those procs seem to have the best price-point.

[grnarrow]

Has anyone here been running a P4 2.4C overclocked at 3.0+? Lots of reviewers seem to have accomplished it fairly easily, and I'm wondering if any of the silence buffs here have done it with fans still running slowly/quietly in their case?

[herosformula]

[i]Hmm.. Then I wonder in what way folding would be so different then from the number-crunching programs they use in the benchmarks. Perhaps the benchmarks are calculating mostly in integers while the folding works with floating point numbers.[/i]

There are floating point numbers and then there are floating point numbers. More than likely, the other tests are using standard 32 bit floating point numbers which can fit into a standard register. If they are optimizing the folding@home units for SSE2, then they are interested in double precision floating point numbers, which are 64 bits wide. In that case, depending on the CPU, the number will not fit into a register, which adds a bit of number gymnastics in order to get each result. Since hardware which is optimized for 32 bit floats may or may not be optimized for 64 bit floats, you can see a larger difference with the folding@home work units versus standard tests.[/i]

[Liquidated]

Don't believe the hype about the springdale.

the spingdale does NOT have performance acceleration (PAT)

the quick and easy about is that unless you have cas 2 super memory (corsair and mushkin come to mind) you cannot use the back door asus and abit are using to speed up the springdale boards.

If you plan on overclocking your 2.4c the moment you relax memory from 1:1 cas2, the 'cheats' will no longer work. Unless you are overclocking why on earth would you spend twice as much on top of the line xms memory as you would on say the VERY capable crucial?

All told, a 2.4c running at 260 fsb on a 5:4 divider (i.e. the memory is only only running at 208 mhz which your basic pc3200 samsung and cruical can do np) on a spingdale and canterwood chip based board from the same manufacturer...

The canterwood board will be faster than the spingdale by 3-4%

honestly though after all this hubbub, does 3-4% mean squat?





as far as I've gleened, The long and the dirty is this.. this board will correct me if I'm blatently wrong, lotta smart cookies here.

spingdales and canterwoods are the same chipset up until the last sorting process. those chips deemed sufficient to move on to be canterwoods undergo one last layer which adds in the PAT features for memory path optimization.

Springdales are the proverbial 99% of a canterwood except their PAT functions are only half implimented... they are missing that last fab layer.

What asus and abit are doing is awakening this half way PAT on the springdale and using it to boost memory performance. These 'cheats' rely on the memory being at low cas speeds, crucial and samsug pc3200 say the lowest their memory will go is cas3.

So, unless you keep the CAS 2 memory at a 1:1 fsb to memory timings, you cannot use this accelleration. This equates into the problem that once you get over 245ish fsb, there's no way you can use the acelleration with even pc 3800 memory.. the springdale has no vestige of PAT and the canterwood, which has pat fully realized, does.

This PAT nets the canterwood about a 3-4% benchmark increase over a springdale... on memory intensive jobs. Meaning in real life terms, who cares unless you are looking for bragging rights?

So there you have it, don't think of the springdale as being as fast or faster than a canterwood, think of it more as the canterwood is not faster in any meaningful way.


debate the merits of the canterwood vs springdale boards on features and cost and you won't be shocked to find out the truth of the matter.

My PC that Ive been twaeking over the past few months after finding this sight is going for a mildly oc'd 2.4c

For cooling the cpu I'm going to use an slk 900u with a 92mm papst. The key about this setup over the alpha pal is that I want the fan overhang to hit the northbridge, since that needs it's own cooling on the canterwood and springdale boards and they are active on the abit and asus.

The problem for me is that the board I originally wanted, the abit IC-7, has the cpu socket turned 90degreees meaning the fan overhang is not over the northbridge.

The springdale based IS-7 on the other hand leaves the cpu socked in it's standard orientation such that the over sized fan will hit the nb. Hence I'm going springdale. I like the lay out of the ic-7 better but oh well.


For barton 2500+ vs. the p4 2.4c, unless you are planning on overclocking the intel chip, stick with the athlon. The barton core is solid and has very strong FPU, the main limitation is that the athlon is near the end of it's life span and doesn't overclock nearly as well as the 2.4c.

I'm going intel this time around and ocing but I'd never suggest a p4c over an athlon throughbred/barton for my friends and family that will never open the case.

and no, don't have the pc done yet, still no time to sit down and sleeve the sparkle.

Cheers!
-Liq

[Jan Kivar]

Don't believe the hype about the springdale.

the spingdale does NOT have performance acceleration (PAT)

At least ABit and Asus have this feature on their Springdale boards. Intel has asked them not to use PAT as the term for it, instead they use names like "game accelerator" and "turbo mode". Here's a nice read (though it's Tom's).

The stuff You said about PAT is true. The performance increase is minimal. You do need the FSB set to 200 (no over/underclocking), and in some boards PAT will work only with two sticks, not four.

Cheers,

Jan

[Liquidated]

I have no problem with asus nor abit and yes they are going against the wishes of intel with the accelleration.

That said, message boards love to start trends (ala the 120mm fortron on this board) some of which are fine others are hype.

The word on the oc/performance related boards was that the springdale was patently faster than the canterwoods because of these back doors.

The truth of the matter is that they can only use the back door under circumstances I'd imagine the majority of this board will never see.

treat PAT on the canterwood as a feature that boosts performance 3-4% at all fsb speeds that you can post. It's always on and works at 3:2 memory as well as 1:1. The real strength of PAT is that relaxing memory settings to achive a higher fsb/clock doesn't degrade performance nearly as much as relaxed timings on a springdale. Again though, that falls within the realm of the benchmark which most on this board simply don;t care about I'd imagine.

You decide for yourself if that's meaningful, I'm not one to say you are wrong =) Just don't think the springdale is faster than the canterwood.

Cheers!
-Liq

[jamoore9]

To build on what Liquidated is saying: if you are the kind of person (like myself) who wants to run stock speeds as tightly as possible in order to maintain cooling and quiet, and to maintain the lifespan of you parts, then "backdoor" PAT is great because you do indeed get that 3-5% -- actually more like 3-8% on the P4P800, depending on the setup -- performance increase.

Bottom line, don't buy a Springdale because you think it has PAT. Buy a springdale because it really does offer comparable performance to a Canterwood, minus 1-4% according to Tom's (ECC memory probably dissipates some of the PAT advantage), for around $100 less! THAT is the Springdale hype: not that it is faster, but that it is fast enough and it is cheap. Besides, and overclocker wouldn't want PAT anyway: who cares about +3% on 5200MB memory bus bandwidth!?

EDIT: And for the record, LOTS of hardware reviews indicate that the P4P800 Springdale at stock settings is often faster than the top performing Canterwood, the P4C800, depending on the test. The Tom's Hardware review above is a case in point. AnandTech came to the same conclusion, as did HardOCP.

[fractal]

I'm going intel this time around and ocing but I'd never suggest a p4c over an athlon throughbred/barton for my friends and family that will never open the case.

If I were spending my money I would go with a barton core athlon. Currently the best value for the money.

If I were making a machine for friends and family that will never open the case -- if I can't talk them into buying a dell, I always have and always will use an intel retail motherboard with an intel retail processor. The cost difference is insignificant compared to what happens when something doesn't work right.

As for bleeding edge. I dunno who is faster in each of the nitch markets. Don't really care either. I can't justify paying 2x as much for 10% faster.