Sempron vs. Athlon tests... where's the Barton?

[ToeJam13]

After reading the latest article on CPU speed tests, I'm left with a bit of a sour taste in my mouth.

The article mainly focused on AMD processors with a core speed of 1.8GHz and with either with 256M of cache (XP & Sempron[754]) versus 512M of cache (Athlon 64)

Although I understand that they wanted an XP processor with the same internal clock as the Sempron, in my humble opinion, would it not have made more sence if they had chosen an XP model with a 512M cache? An Athlon XP/Barton 2500Plus with a 1.83GHz bus and 512M cache would of been a better match than an Athlon XP/Thoroughbred 2200Plus with a 1.80GHz bus and 256M cache.

To go a step further, it would it not of been even better to see a 512M 3100Plus XP going up against a 256M 3100Plus Sempron? People wondering if they should bite the bullet and upgrade from a lastest generation XP mainboard to a Socket 754 mainboard with a Sempron would of enjoyed it. It would have shown how a 256M processor fared against other AMD processors with similar "speed ratings" versus core clock speed.

Personally, I have an ABit AN7 (nForce 2 Ultra 400) mainboard with an older XP processor, and I am considering bouncing up to an Athlon XP/3100Plus. However, if the Sempron pummles it into the ground, I may just bite the bullet and bump up to a Socket754 (or maybe a 939 for that matter).

Comments?

 

[Concillian]

I can't comment on why AT chose the comparison they did, but I'm pretty sure that in most cases the Sempron 3100+ will put the beat down on an XP at any speed.

If all you need to go the XP route is a processor then it might not be a bad idea, though most would buy the mobile and bump it up to 2200 MHz. If you have a really old 266MHz FSB XP and have PC2100 memory, and need to get memory anyway, I'd probably go up to the 3100+ I think in most cases the 3100+ will beat out an XP at any speed. LAME seems to be the notable exception, the AXP seems to be the best LAME processor around (or should that be the LAMEst processor ... *groan*)

 

[lobadobadingdong]

I'm a little confused why they chose the 2200+ instead of the 3000+ barton? apples to oranges.

 

[InlineFive]

Originally posted by: lobadobadingdong
I'm a little confused why they chose the 2200+ instead of the 3000+ barton? apples to oranges.

It does seem that the results are biased to the K8s.

 

[AWhackWhiteBoy]

one would think if you slashed off some of the cache, and lowered the clock frequency, you'd lower the core voltage, and yet the Semprons are 1.6v if i remember correctly. what gives AMD?

 

[Dman877]

I would have liked to see a the 3100+ socket 754 sempron (1.8 ghz) vs a newcastle 2800+ (1.8 ghz) vs a clawhammer @1.8ghz (easy to do with a 3200+ since the lower multipliers are unlocked) vs a barton 2500+ (1.83 ghz) vs a p4 1.8 ghz (just for fun ). Then we would have had cache from 256 on the sempron to 512 on barton/newcastle to 1 meg on the clawhammer...

 

[ToeJam13]

the Sempron 3100+ will put the beat down on an XP at any speed. If all you need to go the XP route is a processor then it might not be a bad idea, though most would buy the mobile and bump it up to 2200 MHz

I dunno. Its a close call up at the top. I have an ABit AN7 (nForce 2 Ultra 400) with 512M mem (DDR400, 2-3-2, dual ch), which is about as nice of a Socket A board you can get. An XP/3200Plus is only $112.

Then, the Sempron/3100Plus is only $129, and Socket 754 boards can now be picked up on the cheap. Write the ABit off as a birthday gift to dad and find something else? Hrm.

 

[ToeJam13]

I would have liked to see ... then we would have had cache from 256 on the sempron to 512 on barton/newcastle to 1 meg on the clawhammer ...

I think Anandtech was trying not to comprare apples to oranges. They were trying to compare Red Delicious to Granny Smith to Jonagold.

Again, my largest complaint of using the 2200Plus is that its not a hotly purchased processor anymore. I know many people getting 3000s and 3200s, but nobody getting anything that old... not unless its off eBay.

 

[Zebo]

Ya he sould have. But though a little interpolation even a 3200+ A-XP would have been destroyed in those tests. This article has an obvious motive to show just how dead the A-XP series is compared to A64/Semiporn. Now go buy one.

 

[ToeJam13]

Hah, maybe. Although I guess I could just get off my butt and look at tests AT has previously done...

XP: AMD XP/3200Plus, DFI NFII mainboard (nForce2 Ultra 400), 2:3:3:6 DDR400 2x256MB dual-channel
SP: Sempron/3100Plus, Gigabyte K8 Triton (nForce3), 2:2:2:6 DDR400 2x256MB single-channel

UT2004 Botmatch 1024x768
XP: 88.9fps
SP: 77.4fps

Gun Metal 1024x768
XP: 36.7fps
SP: 34.3fps

Of course, AnandTech's articles (17.08.2003 and 28.07.2004) aren't exactly helping since their buisness mark and content creation tests don't match, and even those tests aren't 100% reliable since the XP tests didn't say if AA was enabled or not. Blah.