Sparta vs Manila
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
-
- Posts: 247
- Joined: Fri Oct 20, 2006 1:00 am
Sparta vs Manila
From the Peloponnese to the Philippines: I had the opportunity to test the Sempron 3200+ and recent LE-1100 with exactly the same setup. I figured I'd post my results for people who find this kind of stuff interesting.
Here's a head-to-head.
Power Consumption:
As you can see the results are practically the same, except for undervolting where the LE-1100 has a minimal edge over the 3200+. I was surprised to see that the LE-1100 at stock speed let itself be undervolted all the way down to 0.950v. I was even more surprised to see the Sempron 3200+ go even further and run stable at 0.900v. I was accustomed to Semprons 3200+ not going lower than 0.975v. Unfortunately I didn't have time to run longer tests, but the LE-1100 was stable at 0.950v (@ 1.9 Ghz) for 12 hrs under Orthos.
Performance:
Whenever I have the chance I run PCMark05, although I have to admit that I don't always understand the outcome of the tests. For instance here the LE-1100 has a higher CPU score than the 3200+ (fair enough), but why do the system and especially the memory score of the Sempron 3200+ turn out higher when I'm using exactly the same setup? Is it because the LE-1100 is relatively new?
Anyway, the conclusion appears to be that Sempron 3200+'s 'successor', LE-1100, performs a tad better and consumes a tad less (when undervolted).
Here's a head-to-head.
Power Consumption:
As you can see the results are practically the same, except for undervolting where the LE-1100 has a minimal edge over the 3200+. I was surprised to see that the LE-1100 at stock speed let itself be undervolted all the way down to 0.950v. I was even more surprised to see the Sempron 3200+ go even further and run stable at 0.900v. I was accustomed to Semprons 3200+ not going lower than 0.975v. Unfortunately I didn't have time to run longer tests, but the LE-1100 was stable at 0.950v (@ 1.9 Ghz) for 12 hrs under Orthos.
Performance:
Whenever I have the chance I run PCMark05, although I have to admit that I don't always understand the outcome of the tests. For instance here the LE-1100 has a higher CPU score than the 3200+ (fair enough), but why do the system and especially the memory score of the Sempron 3200+ turn out higher when I'm using exactly the same setup? Is it because the LE-1100 is relatively new?
Anyway, the conclusion appears to be that Sempron 3200+'s 'successor', LE-1100, performs a tad better and consumes a tad less (when undervolted).
Very strange. The shrink to 65 nm has to result in less watts and less heat for a Chip that otherwise is pretty similar. Very little gain seems evident.
I'm guessing there may be some bios/motherboard factor that's not letting the Sparta get it's best result----and if that's the case-maybe a week-or a month later,some update unleashes the Sparta and maybe then you get a bit different result.
I'm guessing there may be some bios/motherboard factor that's not letting the Sparta get it's best result----and if that's the case-maybe a week-or a month later,some update unleashes the Sparta and maybe then you get a bit different result.
I think that's being generous. There are basically no differences that could be noticed by someone not doing benchmark tests. on the plus side, amazing idle wattages! did you see that thread where the guy couldn't get consumption under 50W? your max wattage is the same as his lowest wattageAnyway, the conclusion appears to be that Sempron 3200+'s 'successor', LE-1100, performs a tad better and consumes a tad less (when undervolted).
-
- Posts: 247
- Joined: Fri Oct 20, 2006 1:00 am
I was quite amazed when reading that. That guy should definitely spend more time on SPCR.jaganath wrote: on the plus side, amazing idle wattages! did you see that thread where the guy couldn't get consumption under 50W? your max wattage is the same as his lowest wattage
Yes, I also think this might be the case. On Gigabyte's support site it says this:I'm guessing there may be some bios/motherboard factor that's not letting the Sparta get it's best result----and if that's the case-maybe a week-or a month later,some update unleashes the Sparta and maybe then you get a bit different result.
F5: Support AM2 G1 stepping CPU
F6: Update CPU ID (Support AMD Athlonâ„¢ 64 X2 BE-2300 &BE-2350 CPU)
F7: Fix CPU compatibility issues (Sometimes booting failed after reset)
F8: Update CPU ID (Support AM2 G1/G2-Stepping CPU)
F8 is 'update' but not support. We'll notice soon enough as I'll be working with these chips for a while. I'll report if anything changes.
-
- Posts: 192
- Joined: Sun Apr 08, 2007 5:09 pm
Don't assume that as they shrink process that they HAVE to reduce power.
When you work the variables, you can optimize for power, performance or a mix of both. Apparently they went for performance.
My Athlon 5000+ system pulls ~47W, which is not that much more considering twice the ram and a faster proc.
When you work the variables, you can optimize for power, performance or a mix of both. Apparently they went for performance.
My Athlon 5000+ system pulls ~47W, which is not that much more considering twice the ram and a faster proc.
35W manila core Sempron 3200+? It's very difficult to get one, and it's even rarer than X2 3800+ 35W. I guess the mentioned Sempron 3200+ would be 59W TDP, and not 35W.yama wrote:I just wanted to mention that while the sparta are 65nm compared to the manila 90nm, it has a 45W tdp rather than a 35w for the manila. someone needs to compare this and see what one performs with less power.
But I was a bit surprised by the no power draw difference between 90nm & 65nm.
You're right - I was looking at the wrong list.
Its 62W for the normal manila but that doesnt really help solve the issue.
I've been planning to build a low power sempron system so the LE-1100 results were a letdown. From what Palin showed, there is pretty much no difference in the new core. I was hoping for something a little more. Now I'm not so sure what I want to do.
That E2140 is suddenly looking better... I hear it can run passive cooling on stock heatsink, and low power on full load as well as overclock above 3Ghz. It pretty much has something for everybody.
Its 62W for the normal manila but that doesnt really help solve the issue.
I've been planning to build a low power sempron system so the LE-1100 results were a letdown. From what Palin showed, there is pretty much no difference in the new core. I was hoping for something a little more. Now I'm not so sure what I want to do.
That E2140 is suddenly looking better... I hear it can run passive cooling on stock heatsink, and low power on full load as well as overclock above 3Ghz. It pretty much has something for everybody.
Because you take single core sempron into consideration, I guess you've no demanding usage. I suggest you save money to buy a low-end sempron, athlon64 3000/3200+ instead. If you've no need, why bother it? If you've need in the future, you'll get a better cpu at the same or cheaper price.
There's no need to buy a computer in a hurry, their price will go down gradually.
There's no need to buy a computer in a hurry, their price will go down gradually.
Re: Sparta vs Manila
I guess it's due to the memory divider thing. At 1.8Ghz, it can use a divider of 7 for a memory frequency of 257Mhz. At 1.9Ghz however, divider 7 would give 271Mhz - I'm sure any decent (or even not decent...) DDR2-266 memory can do that without trouble but technically this is still overclocking. Thus the bios probably falls back to use divider 8 - for a frequency of only 237.5Mhz.Palindroman wrote: Whenever I have the chance I run PCMark05, although I have to admit that I don't always understand the outcome of the tests. For instance here the LE-1100 has a higher CPU score than the 3200+ (fair enough), but why do the system and especially the memory score of the Sempron 3200+ turn out higher when I'm using exactly the same setup? Is it because the LE-1100 is relatively new?
And I guess it could have quite some influence on the system score since you're using a slow single-channel setup to begin with. Try changing the memory divider...