Which is faster Athlon XP 2400+ or Sempron 2800+ 754?

[Zap]

Originally posted by: ElFenix
i've just found out that my s754 board doesn't fit in the case i wanted to use it in...i suppose i could do the measurements without putting it in a case, but that makes it a bit more difficult to return to fry's (opened and used).

Not a problem for returns as long as you return it within their 2 week period. Just bring it back (opened and used even) and say the exact truth, that "it didn't fit my case properly." They will give you a full refund.

Originally posted by: Calin
It is just enough for what I need to do with it, and I didn't want to upgrade it until nowadays. Nothing I could have bought at that time would have an easy upgrade path.

Thank you. Calin has a "real world" example of what I'm trying to tell people - unless you really do intend to upgrade soon, there is no point in paying extra for an "upgrade path" that won't get used.

Originally posted by: ta8689
But when you want to get a new processor... The extra 100 might keep you from having no option other than to upgrade your mobo, cpu, and vid card if you want to upgrade at all.

You can get PCIe socket 754 boards. The $100 savings can go towards a socket AM2 motherboard. The money for an upgraded socket 939 CPU can go towards a socket AM2 CPU. Sell off socket 754 CPU, motherboard and DDR SDRAM, and use the proceeds to buy some DDR2 SDRAM.

The look on your online buddy's emoticon when you tell him you have a new AM2 setup while he's using a year old socket 939 motherboard... priceless. :Q

 

[Zap]

Actually, was just thinking... I know people who like to get the best that they can afford, but their reasoning is NOT that they have an upgrade path. They want to increase the time between system upgrades because they don't like having to mess with changing hardware and reinstalling Windows. So what have we learned?

Some people just have plenty of money, so they may as well get socket 939.

Some people fully expect to upgrade again in the near future, so they may as well get socket 939.

Some people don't want to have to mess with their systems for a long time. Probably doesn't matter which platform they get as long as they get near top-of-the-line to maximize time before having to upgrade again, but socket 939 may be better.

Some people don't like wasting money so there's nothing wrong with socket 754 for them.

Some people like having good bang-for-the-buck, so there's nothing wrong with socket 754 for them.

Some people absolutely do not have the money for anything better, so there's nothing wrong with socket 754 for them.

Some people absolutely do not use their computers for anything more stressful than typing up emails, so there's nothing wrong with socket 754 for them.

 

[bigsnyder]

This article touches on this issue. It would be nice to see some updated benchmarks for the AthlonXP that includes
the new 7800GS and and ATI's best AGP offerings. Obviously a boarder range of benchmarks than just HL2 would be helpful.

C Snyder

 

[RobsTV]

Originally posted by: ElFenix
tbred 2ghz/333 (xp 2400+/sempron 2800+)
pc mark 04: 3027
pc mark 05: 2358
3dmark2001: 10257
q3: 169 fps

i'll be back with the 754 scores in a bit

That is terrible for Socket A gaming.
I sure hope that type of system is NOT what some are saying is a normal Socket A gaming rig.
Heck, my old overclocked GF3 Ti500 or ATi 8500 with XP2000+ beats that.
That is not much better than todays onboard
(Jetway A210 Pro xpress200 scores 8000 3dm2k1se)
Upgrade video, and be done with it.
Even an ATi 9700 Pro (for about $70) would nearly double performance. (17,000 3dm2k1se).
Probably could pick up an ATi 9800 Pro for about the same price.
The old socket A would then be faster than a new socket 939 running a $100 ATi x700 PCIe card.

 

[PingSpike]

Originally posted by: Zap
I guess my whole point for this mental masturbation is that there isn't as much difference in performance as people are hoping/wanting between socket 754 Semprons and socket 939 Venice.

I agree. Ever since 939 came out, everyone has been crapping on 754. Don't buy that, its a dead upgrade path they say. No dual channel! No dual cores. I own a 939 venice myself, and am quite happy with it. But, if you're planning on slapping together a budget box and aren't that concerned about upgrability in the immediate future, 754 is a bargain platform. I built a beastly little overclocking machine for my wife for a lot less that performs almost as well. The semprons are cheap! The motherboards are cheap!

Originally posted by: Zap
Basically the same except for a higher FSB.

Man, haven't looked at socket A stuff for a while, seems like the pricing has gone up with the inflated Sempron ratings while the MHz has stayed down.

...which brings me to another point. I upgraded my machine to socket 939 and a venice. I had a barton mobile at 2.3ghz and a NF7 left over that could go into my wife's machine. But I'd been watching the socket A stuff steadily climb in price, and seen a few forum members praising the cheap performance combo of socket 754 semprons while being drowned out by the "socket 754 is a dead end! no dual channel!" crew. I checked some benchmarks (its HARD to find a lot of direct comparisons between socket A and the socket 754 semprons btw, but xbit had a pretty good feature on it) and took a look at the market on used parts.

I sold my socket A stuff, and bought a chaintech VNF3-250 and a palermo 256k cache 2800+. Cost me $10-15 when all was said and done, and I could have made money on the trade but I just priced the parts to sell instead of trying to get the maximum cash for them. I got this beast purring away at 2.33ghz on stock voltage. It doesn't matter if that barton would have beaten the sempron at its stock speeds...it'll get soundly beaten by it at the same clock.

 

[bigsnyder]

@PingSpike

Could you post a link to that xBit article? I am having a hard time finding it. Sounds like
an interesting read.

C Snyder

 

[PingSpike]

Originally posted by: bigsnyder
@PingSpike

Could you post a link to that xBit article? I am having a hard time finding it. Sounds like
an interesting read.

C Snyder

Sure. Its not as direct of a comparison as I would have liked (like I said, hard to fine benches putting these two head to head) but it does compare a sempron on socket A to one of the weaker 754 semprons with only 128k of cache. You have to figure the barton would do a little better on socket A, and the 2800+ would do better on s754. Unfortunately that was the best/closet head to head review I could find.

It was still kind of fun seeing the p4 3.4 EE getting beaten by an overclocked sempron in superpi.

I really should have done some before and after benches myself, but I'd already torn my machine apart and didn't have a graphics card to do the job justice at the time.

There was a thread I ran across that I can't find right now called either the socket A challenge or the socket 754 challenge or something. It was on another forum I don't frequent, found it through google...where socket A users were invited to compare their mobile bartons to the OPs sempron setup. It was a great read, and had a lot of the barton users surprised.

 

[Zap]

That article's an interesting read. So the socket A and 754 versions of the same "rated" Sempron at 2600+ speeds is pretty close - nice give and take on various synthetic benchmarks. The PCMark memory score was much better on socket 754 than A, probably from the integrated memory controller. The encoding was 2 for 2, but interesting to see that the two which socket 754 lost looked like a big margin. I'm wondering if that's because those particular two don't fit into 128k cache and the socket 754 2600+ was the only CPU in those tests under 256k cache. Perhaps this is also the reason the 128k cache Celerons were so weak.

Now how about actual game benchmarks? The socket 754 Sempron spanked socket A even though the socket A version had twice the cache and 230MHz more core speed.

Overclocked? What's a common overclock for desktop socket A CPUs? Well, if we were to include the better overclocking mobile CPUs that would be around 2.4GHz. Well, that's similar to socket 754 Semprons, and if at a deficit they can perform better in games, what happens when there is no MHz deficit and no cache deficit (using the 256k cache Palermos)?

C'mon, somebody out there must have done some benchies!!!

 

[ElFenix]

ok ok i'm gutting my k7s5a system... it just sits in the closet anyway...

when i'm done i can stick my 9800 back into my mobile barton and give the tests a whirl there at 2.3 ghz (rev 1.06 of the k7n8x doesn't go to 200 mhz )

 

[kmmatney]

Originally posted by: hurtstotalktoyou
A little while ago I moved from an Athlon XP 2400+ & 1GB PC2100 to a Sempron 64 2800+ & 512MB PC3200. Even with less RAM, the Sempron was most definitely faster. Factor in my 50% overclock, and it blows my old 2400+ out of the water.

Pretty much the exact same upgrade I did. Factor in X800 GTO2, unlocked and OC'ed and you have an amazing bang-for-the-buck gaming system. Even at stock, the Sempron 2800+ blew away my old Athlon XP 2400+.

Outpost has a deal for a Sempron 3100+ cpu and ECS motherboard:

http://shop3.outpost.com/product/4705559

This is a desirable Sempron with 256K of cache. Even if its a Paris core, still a great deal. If its a Palermo core, even better.

 

[Zap]

Someone's hitting 2.7GHz on his 3100+ Palermo with a Biostar Geforce\Tforce6100 board (he didn't specify which one). He says not Prime stable, but can loop game benchmarks at that speed. :Q thread

 

[RobsTV]

An interesting comparison between Opteron 146 and AXP 3200+

http://www.driverheaven.net/reviews/Opteron/7%20pcmark04.htm

Looks like the AXP keeps up rather well in most tests, (staying within 10% on most, and even winning some tests), so if those here with Sempron 754's are "blowing away" AXP's, then they must also be "blowing away" Opteron's, or smoking something real good.

Point is, we all know 754 and 939 are faster than Socket A.
How much faster? Not much. Nowhere close to the hype. In fact, in some apps or systems, a decent socket A system will be faster. Worth the money to upgrade for marginal results? Leave that up to the buyers wallet. But if they base there purchase on posts here, they will be in for a sad sorry experience. Socket A will NOT be blown away by even socket 939 under most conditions. Mild upgrade path at best.

For comparison, going from a slow Socket A to a fast Socket A is about the best type of speed increase you could expect to get going from a fast Socket A to a fast Opteron. 754 is of course worse than 939.

Signed, an Opteron upgrader that bought into the hype, (6 machines running 2.7GHz-3GHz, or faster than most here run there 939's) and was shocked to see such a slight increase in speed from AXP @ 2100MHz.

If looking to purchase a system, it would be foolish to get socket A.
754 with PCIe is the best bang for the buck, if looking for new.
But upgraders should wait.

 

[PingSpike]

RobsTV brings up some good points, and going back and looking at the very benchmarks *I* posted I think I was probably a little overly enthusiastic with my recommendations. I got a free upgrade by selling off the socket A stuff and am still very happy with my decision...but its true that we aren't seeing wholesale destruction of the XP at the hands of the sempron here. I mainly concern myself with gaming benchmarks since they're the only ones that count to me, I've found pretty much every processor 1ghz or better adequete for almost everything else I do.

More the point is, we aren't really seeing that "oh wow!" increase between processor cycles anymore. Folks on northwoods and bartons still do well on the latest games...and when did those cpus originally come out? Is there anything BUT a mild upgrade path when it comes to processors these days?

Everyone says that dual core will provide a big benefit...but until I see software take advantage of it on a much wider scale I just can't see a lot of bang for buck in going that route yet either.

 

[maluckey]

Looks like the AXP keeps up rather well in most tests, (staying within 10% on most, and even winning some tests), so if those here with Sempron 754's are "blowing away" AXP's, then they must also be "blowing away" Opteron's, or smoking something real good.

I still haven't seen any other first hand information to compare with posted on this thread. I am truly curious as to how others sytems fare in comparison. With so many members, we should be able to post tens of different types of information to compare with. Let's all submit a few results that are typical of systems found around the web.

 

[ElFenix]

from driverheaven:

The Opteron is running ddr266 (dual channel)

wtf? why? part of the benefit of 939 is extra memory bandwidth.

 

[abhaxus]

going from an XP-M 2400+ overclocked to 2333mhz to an athlon X2 3800+ resulted in a HUGE difference in playability in all games, including single threaded ones. before i started overclocking, and before i updated Q4 to the 1.0.5 beta patch, I went from 48fps average in a custom timedemo at 800x600 lowest settings, to a steady 65 fps in the same time demo at 1024x768 with 4x AA and max settings all around. of course with the dual core patch the numbers are higher, around 80fps.

the cheapest/best upgrade path is a new 939 chip + the asrock board with pcie and agp. you get to use your current card (for me, I had an X800XL agp card) and then upgrade later on.

plus the A64 based chips overclock the best. 2333 was patheticly slow at q4. even at 2 ghz my A64 was faster, and now with it at 2500 I get an even more noticeable speed boost in games like HL2 which are not multithreaded.

 

[Zap]

Originally posted by: ElFenix
from driverheaven:

The Opteron is running ddr266 (dual channel)

wtf? why? part of the benefit of 939 is extra memory bandwidth.

That makes it suspect. Hmmm.

Originally posted by: PingSpike
I got a free upgrade by selling off the socket A stuff and am still very happy with my decision...but its true that we aren't seeing wholesale destruction of the XP at the hands of the sempron here. I mainly concern myself with gaming benchmarks since they're the only ones that count to me,

More the point is, we aren't really seeing that "oh wow!" increase between processor cycles anymore. Folks on northwoods and bartons still do well on the latest games...

The Sempron did better on all the game stuff but then again that was only a handful of games.

For purchasing a new system the Sempron is the only choice IMO (between 754 and A). When upgrading, then you have to take a lot of other things into consideration. One thing is that the "good" socket A CPUs are getting tough to find and the prices seem to be going up.

EDIT: Forgot to mention that my gaming rig is a Northwood a few years old. No HT but running at 3.33GHz with a 6800GT. Works beautifully.

 

[carlosd]

Originally posted by: RobsTV
An interesting comparison between Opteron 146 and AXP 3200+

http://www.driverheaven.net/reviews/Opteron/7%20pcmark04.htm

Looks like the AXP keeps up rather well in most tests, (staying within 10% on most, and even winning some tests), so if those here with Sempron 754's are "blowing away" AXP's, then they must also be "blowing away" Opteron's, or smoking something real good.

Point is, we all know 754 and 939 are faster than Socket A.
How much faster? Not much. Nowhere close to the hype. In fact, in some apps or systems, a decent socket A system will be faster. Worth the money to upgrade for marginal results? Leave that up to the buyers wallet. But if they base there purchase on posts here, they will be in for a sad sorry experience. Socket A will NOT be blown away by even socket 939 under most conditions. Mild upgrade path at best.

For comparison, going from a slow Socket A to a fast Socket A is about the best type of speed increase you could expect to get going from a fast Socket A to a fast Opteron. 754 is of course worse than 939.

Signed, an Opteron upgrader that bought into the hype, (6 machines running 2.7GHz-3GHz, or faster than most here run there 939's) and was shocked to see such a slight increase in speed from AXP @ 2100MHz.

If looking to purchase a system, it would be foolish to get socket A.
754 with PCIe is the best bang for the buck, if looking for new.
But upgraders should wait.

As can you can see from some comparisons, the Semprom 754 and Athlon 939CPUS at same clock speeds show difference of 1-5% in general apps and 8-15% in games (given you have fast GPU) running at same clockspeeds. Those are only PCMARK benchies and you can't draw absolute conclusions by only looking at them. In general terms (You draw that conclusion makin various types of tests, not only PCMARK) The opteron/A64/Semprom754 will blow away any AXP running at the same clockspeeds.
An athlon XP WILL NEVER BE FASTER THAN AN AMD64 BASED CPU RUNNING AT SAME SPEED IN REAL APPS.

Marginal results are 15-30% of performance increase even if the Semprom 754 is running a much lower clockspeeds?. I don't think so. For me 30% running 400MHz slower is much much faster.(not forgetting the great OC potential of the Palermo) And socket 754/939 do blow away Socket A in everything, you really need to see much more tests.


As you can see the difference between Semprom 754(128K cache) And A64 (512K cache) running at same clockspeeds is marginaly 1-5%, except in games.
http://www.xbitlabs.com/articles/cpu/display/sempron-3000_7.html

Palermo at 2.6GHz. No AXP will near of beating that. (pure performance, not talking about clockpeeds)
http://www.xbitlabs.com/articles/cpu/display/sempron-3400_5.html

Also, influence of L2 cache on performance on AMD64 CPUs. The effect seems to be dismal in general performance, except in games and cache dependat apps.
http://www.xbitlabs.com/articles/cpu/display/sempron-2600_4.html

 

[maluckey]

Here are a few other Sempron related benches.

Sempron Paris, versus Sempron T-Bred versus Barton Athlon (desktop)

Barton (mobile) 2500 versus Sempron 2600

In the second link, I would like to point out that likely no enthusiast would run the T-Bred Sempron any more than a Barton XP-M owner runs at default freq or voltages. This is just linked to to give perspective.

 

[carlosd]

Originally posted by: maluckey
Barton (mobile) 2500 versus Sempron 2600

The first link, what we all know the 3100+ whiping the Higher clocked barton 3200+

In the second link this is a semprom 2600+socket A t-bred they made the comparison to. We are talking about AXP cores (tbred, barton) Vs. Socket 754 cores (palermo, paris)

 

[Fox5]

Originally posted by: RobsTV
An interesting comparison between Opteron 146 and AXP 3200+

http://www.driverheaven.net/reviews/Opteron/7%20pcmark04.htm

Looks like the AXP keeps up rather well in most tests, (staying within 10% on most, and even winning some tests), so if those here with Sempron 754's are "blowing away" AXP's, then they must also be "blowing away" Opteron's, or smoking something real good.

Point is, we all know 754 and 939 are faster than Socket A.
How much faster? Not much. Nowhere close to the hype. In fact, in some apps or systems, a decent socket A system will be faster. Worth the money to upgrade for marginal results? Leave that up to the buyers wallet. But if they base there purchase on posts here, they will be in for a sad sorry experience. Socket A will NOT be blown away by even socket 939 under most conditions. Mild upgrade path at best.

For comparison, going from a slow Socket A to a fast Socket A is about the best type of speed increase you could expect to get going from a fast Socket A to a fast Opteron. 754 is of course worse than 939.

Signed, an Opteron upgrader that bought into the hype, (6 machines running 2.7GHz-3GHz, or faster than most here run there 939's) and was shocked to see such a slight increase in speed from AXP @ 2100MHz.

If looking to purchase a system, it would be foolish to get socket A.
754 with PCIe is the best bang for the buck, if looking for new.
But upgraders should wait.

Just to point out, an FX-60 (or is it 61? the dual core), on average has around 1.5x to 2x the performance of an XP 3200+, depending on the level of multithreading.
Oh, and that opteron is using PC2100 memory, largely negating the decreased latency brought about by the memory controller.

 

[carlosd]

OK robs TV if you are Lazy to look at the THG charts I've made it for you and have compared clock for clock performace of Barton and Semprom 754 (palermo) CPUs both at 2GHz. I discarted the synthetic benchies (except 3Dmark) because they are useless.

I converted the minutes to seconds.

So at same clockspeeds:


Video
Units (Seconds)

Test...............Sempy 2GHz............Barton 2Ghz ..........Diff %.....Comments

DivX6.....................518............................682...........31,66%..........Huge!!
Pinnacle St.9.........142............................159............11,97%
CloneDVD.............710............................798............12,39%
Main concept.........191............................259............35,60%.........Huge!!
Media enc. St........196............................229............16,84%

Audio
Units (seconds)

Test...............Sempy 2GHz............Barton 2Ghz ..........Diff %.....Comments
Ogg 1.1..................245............................306.............24,90%
Lame MP3 ENC......265............................341.............28,68%........Huge!!


3DSmax.................172............................97................14,53%


Gaming
Units(FPSs, except 3dmark)

Test...............Sempy 2GHz..............Barton 2Ghz ..........Diff %.........Comments

3DMARK05...........4935..........................4065.............17,63%
DOOM3 1024........85,4...........................59,3.............30,56%.......Huge!!, but still synt.
Farcry...................161,8.........................151,6............6,30%........OK, it seems it's not so CPU bounded
UT2004.................131,2.........................80,9..............38,34%......Hugeee!!


Multitasking
Units (secs)

Test................Sempy 2GHz..............Barton 2Ghz ..........Diff %....Comments
Multitasking I........218...........................297...............36,24%......Huge!!
Multitasking II.......683...........................1061.............55,34%......HUGGEE!!!

All the test shows the semprom whipping the barton.
Where is the only 10% in diff you were talking about?? laughable it seems. The difference based on multiple tests between Barton and Semprom 754 at same clockspeeds is huge!!!, can't imagine after the overclocking given that palermo core can reach 2.5-2.6GHz, you would need a 3.2 GHz Barton to reach that, oh...wait that's imposible!

PD: Never thouhg those THG charts were going to be that usefull :laugh:

 

[carlosd]

Originally posted by: Fox5

Oh, and that opteron is using PC2100 memory

O I haven't seen that, but now that I,ve realized, Rob let me tell you: that comparison is absolutely useless. Nobody will be so stupid to use an opteron with DDR266 memory.

 

[RobsTV]

Most missed the whole point.

The top of the line average A64 single core is Opteron, then comes A64, then comes Sempron.
So even if an Opteron is using slow memory, it still has 1 meg cache, and should still be faster than Sempron.

I'm really suprised that not one of you experts caught that the Opteron tested was a socket 940......

Still, is a slow Opteron w/1meg cache as fast as a Sempron 754?

Just showing another point of view after real world testing.
What good is a one sided discussion?
My money was not spent, (always resell), so I wasted nothing upgrading to 939 Opteron's.
Had I spent the cash, I too would defend my purchase. It's only human nature.

 

[carlosd]

Originally posted by: RobsTV
Most missed the whole point.

The top of the line average A64 is Opteron, then comes A64, then comes Sempron.
So even if an Opteron is using slow memory, it still has 1 meg cache, and should still be faster than Sempron.

I'm really suprised that not one of you caught that the Opteron tested as a socket 940......

Still, is a slow Opteron w/1meg cache as fast as a Sempron 754?

It depends on the app. Again you are just showing PCMARK bechmark... it is just not enough to say that a CPU is faster than another. I am sure a opteron with dual DDR400 will beat a semprom running at the same clockspeed in everything, didn't you see the xbitlabs links (i see you are lazy) that shows the influence of cache in AMD64 CPUs?? There you will have a picture how 1MB of chache influences the perfromance compared to 128K or 256K AMD64 CPUs like semprom.

No one missed the point. You pointed out a useless test comparing an opteron using DDR2266, and you drawed general conlcusions based ONLY IN ONE SINGLE SYNTHETIC TEST which I have said many times are useless. LOOK AT THE REAL APPS. AND YOU WILL SEE OPTERON KICKING AXP's ASS. That's it.

You have to use some aristothelic logic reasoning.

Semprom 754 blows away Athlon XP clock for clock.

Opteron is a faster CPU than Semprom 754.

So Opteron will blow away the Athlon XP clock for clock.