I recently made the jump that I suppose alot of you out there are considering. I went from my K8 (4600 X2 Brisbane) to a 720 Black Edition PhII. I thought that since I was wondering how the two really compare, that I would do a fairly comprehensive clock for clock test between the two.
First off, I used my sig rig for all of these tests. And while I did my best to make these benchmarks accurate, I didn't run as many tests of each benchmark as some sites do, nor did I optimize my system for each run. I ran it as it normally is with all background services up and running. So the scores that you see here may not match exactly what you find elsewhere, but it WILL give you a chance to see exactly how much better the PhII is in everyday circumstances, which is really my intent here.
Here are the specs for each setting:
My 4600 was benched at totally stock speeds, and the rams default DDR800 5-5-5-15 2T timings. This will be the blue bar on all of the following graphs.
I ran my 4600 at its OC as well. It was running at 3GHz even (250x12) and the ram was OC'd to DDR1000 at the same timings. This will represent a fairly typical K8 OC (most usually fall between 2.8-3.3GHz, so I figured 3GHz would be a good representation for most people). This is the orange bar.
I actually underclocked my 720 BE as soon as I put it in (actually tough to do, since I wanted to see what it would do ASAP). I lowered the multi to 12x, so that I could do an exact clock for clock comparison with the 4600. This means that if you have a K8 that is different than mine, you should still be able to come up with about how it will perform with a similarly clocked PhII. This is the yellow bar.
The green bar is the 720 at stock 2.8GHz.
The purple bar is the 720 BE at a marginal OC (having issues OCin with my board... its taking way to much voltage to get to where I did). It is running at 3.6GHz, and this was done simply with the multi raised to 18x, and increasing the voltage. This should let you know at LEAST where you should be able to get the 720 BE to, but you'll likely get higher.
While I needed large amounts of vcore (1.55v)to get this OC, the temps were still great. I never saw it break 39c (I do have a cool room, but I would willing to guess that the temps were at least within a few degrees c of the actual temp, as this is the program that came with the board, and actually reported my brisbane right). The Xiggy is holding it's own, I'm impressed.
Ok, here we go. The first round of benches is everyone's favorite, 3DMark06. Notice how even the underclocked 720 BE easily beats the OCed K8 in the CPU score, and how the OC'd 720 trashes everything else in that category. The total score of the OC'd 720 isn't much higher, because it is being bottlenecked by my 9800GT (before, it was the other way around, as my score jumped by a huge margin when I OC'd my 4600).
Next in line is 3DMark05. You can notice the same trends here.
7Zip is an archival program that is similar to WinRAR, but free. One thing to point out, is that while this program is multicore aware... it only does it in even core increments. So that means that the 720 BE is only using two of its cores here, which is why the scores are fairly close. Still, notice that the 720 at 2.8GHz beats the OC'd K8 at 3GHz.
This benchmark I find interesting to watch, as it DOES make use of all cores present, and as you can see, this gives the 720 a large advantage here.
Note: I made the mistake of using the 720 with a lowered NB speed at first, as in this test (and one other, which you will see later on) I ran the 720 at a reduced NB clock on accident, so I think this is why you see the same score for the Open GL tests at 2.4 and 2.8GHz. This would help explain the excellent performance of the OC'd K8, as it has an extra 25% of memory bandwidth. The OC'd 720 is using the NB at full speed, but DDR800. Impressive.
This one is one that I found on the web. It does 3 tests, and then simply tells you how long it took to complete them. Judging by the results, I don't think that this program is very multicore aware.
Passmark7 is a great program, it benches the complete system, and will give you a total rating, and then detailed info in each area. The downside, is that it appears not to be multicore, but it is still a good comparison tool.
This is the total system score, and the following will be the CPU and memory results, since those are what interest us here the most (everything else being constant in my system).
Here is the first round of CPU scores (had to break them up to be readable, as some scores are large, and others small). AMD seems to have made a significant improvement in how the cache and instructions operate (look at the string sorting test, the OC'd K8 is overtaken by even the underclocked 720).
And the second round of results for the CPU. Notice how well the 720 did in the physics test. I think this is a result of the large cache increase, but dont have anything to back up that statement atm.
Here is passmarks memory test (this is the other test that the low NB speed is apparent for the underclock and stock clocked 720). The part that I find most impressive here, is that the 720 when OC'd has a large lead on the 3GHz K8, even though the ram is running 200MHz slower. AMD has definitely improved the IMC on the AM3's. I'll try to do more on this later with the 720 using DDR1000.
POVRay is another rendering program, similar to Cinebench (if you want to bench your own rig, I'd suggest Cinebench, the interface is better, and just simply works better). Again, the 720 clearly has a clock for clock advantage here.
The top test is pixels per second (higher is better), while the other is the time (in seconds) it took to render the benchmark file, so lower is better here.
And finally, here is the classic Whetstone. It seems like most sites that review will use this, so you can compare this to lots of other scores if you'd like.
Well, that just about wraps things up for this test. If any of you would like to test your own rigs, you can download any of the same files and run them yourselves.
http://www.7-zip.org/
http://www.maxon.net/pages/download/cinebench_e.html
http://www.futuremark.com/benchmarks.../introduction/
http://www.softpedia.com/progDownloa...oad-10213.html
http://www.snapfiles.com/get/povray.html
http://majorgeeks.com/Whetstone_d3928.html
http://www.passmark.com/products/pt.htm
You might like to know that I got quite a few of these from Bit-Tech.net. They actually used alot of the same tests, so you can check them out HERE. Just remember that my system is totally different, which is why the scores differ so much at times.
Also, I don't claim this to be a perfect review, but if you have constructive criticism, then I'll see what I can do to make things better.
And I'll let you know now that I have a COD4 test coming up, but I didn't have it finished in time. And I might be able to find some other titles to test as well, but thats not for sure yet.
Once again, I did this because simply, I would have LOVED one of these while I was making the decision to buy mine, and figured that you (and random people searching the web) might feel the same way.
Cheers!
Edited by logan - 3/2/09 at 9:31am
First off, I used my sig rig for all of these tests. And while I did my best to make these benchmarks accurate, I didn't run as many tests of each benchmark as some sites do, nor did I optimize my system for each run. I ran it as it normally is with all background services up and running. So the scores that you see here may not match exactly what you find elsewhere, but it WILL give you a chance to see exactly how much better the PhII is in everyday circumstances, which is really my intent here.
Here are the specs for each setting:
My 4600 was benched at totally stock speeds, and the rams default DDR800 5-5-5-15 2T timings. This will be the blue bar on all of the following graphs.
I ran my 4600 at its OC as well. It was running at 3GHz even (250x12) and the ram was OC'd to DDR1000 at the same timings. This will represent a fairly typical K8 OC (most usually fall between 2.8-3.3GHz, so I figured 3GHz would be a good representation for most people). This is the orange bar.
I actually underclocked my 720 BE as soon as I put it in (actually tough to do, since I wanted to see what it would do ASAP). I lowered the multi to 12x, so that I could do an exact clock for clock comparison with the 4600. This means that if you have a K8 that is different than mine, you should still be able to come up with about how it will perform with a similarly clocked PhII. This is the yellow bar.
The green bar is the 720 at stock 2.8GHz.
The purple bar is the 720 BE at a marginal OC (having issues OCin with my board... its taking way to much voltage to get to where I did). It is running at 3.6GHz, and this was done simply with the multi raised to 18x, and increasing the voltage. This should let you know at LEAST where you should be able to get the 720 BE to, but you'll likely get higher.
While I needed large amounts of vcore (1.55v)to get this OC, the temps were still great. I never saw it break 39c (I do have a cool room, but I would willing to guess that the temps were at least within a few degrees c of the actual temp, as this is the program that came with the board, and actually reported my brisbane right). The Xiggy is holding it's own, I'm impressed.
Ok, here we go. The first round of benches is everyone's favorite, 3DMark06. Notice how even the underclocked 720 BE easily beats the OCed K8 in the CPU score, and how the OC'd 720 trashes everything else in that category. The total score of the OC'd 720 isn't much higher, because it is being bottlenecked by my 9800GT (before, it was the other way around, as my score jumped by a huge margin when I OC'd my 4600).
Next in line is 3DMark05. You can notice the same trends here.
7Zip is an archival program that is similar to WinRAR, but free. One thing to point out, is that while this program is multicore aware... it only does it in even core increments. So that means that the 720 BE is only using two of its cores here, which is why the scores are fairly close. Still, notice that the 720 at 2.8GHz beats the OC'd K8 at 3GHz.
This benchmark I find interesting to watch, as it DOES make use of all cores present, and as you can see, this gives the 720 a large advantage here.
Note: I made the mistake of using the 720 with a lowered NB speed at first, as in this test (and one other, which you will see later on) I ran the 720 at a reduced NB clock on accident, so I think this is why you see the same score for the Open GL tests at 2.4 and 2.8GHz. This would help explain the excellent performance of the OC'd K8, as it has an extra 25% of memory bandwidth. The OC'd 720 is using the NB at full speed, but DDR800. Impressive.
This one is one that I found on the web. It does 3 tests, and then simply tells you how long it took to complete them. Judging by the results, I don't think that this program is very multicore aware.
Passmark7 is a great program, it benches the complete system, and will give you a total rating, and then detailed info in each area. The downside, is that it appears not to be multicore, but it is still a good comparison tool.
This is the total system score, and the following will be the CPU and memory results, since those are what interest us here the most (everything else being constant in my system).
Here is the first round of CPU scores (had to break them up to be readable, as some scores are large, and others small). AMD seems to have made a significant improvement in how the cache and instructions operate (look at the string sorting test, the OC'd K8 is overtaken by even the underclocked 720).
And the second round of results for the CPU. Notice how well the 720 did in the physics test. I think this is a result of the large cache increase, but dont have anything to back up that statement atm.
Here is passmarks memory test (this is the other test that the low NB speed is apparent for the underclock and stock clocked 720). The part that I find most impressive here, is that the 720 when OC'd has a large lead on the 3GHz K8, even though the ram is running 200MHz slower. AMD has definitely improved the IMC on the AM3's. I'll try to do more on this later with the 720 using DDR1000.
POVRay is another rendering program, similar to Cinebench (if you want to bench your own rig, I'd suggest Cinebench, the interface is better, and just simply works better). Again, the 720 clearly has a clock for clock advantage here.
The top test is pixels per second (higher is better), while the other is the time (in seconds) it took to render the benchmark file, so lower is better here.
And finally, here is the classic Whetstone. It seems like most sites that review will use this, so you can compare this to lots of other scores if you'd like.
Well, that just about wraps things up for this test. If any of you would like to test your own rigs, you can download any of the same files and run them yourselves.
http://www.7-zip.org/
http://www.maxon.net/pages/download/cinebench_e.html
http://www.futuremark.com/benchmarks.../introduction/
http://www.softpedia.com/progDownloa...oad-10213.html
http://www.snapfiles.com/get/povray.html
http://majorgeeks.com/Whetstone_d3928.html
http://www.passmark.com/products/pt.htm
You might like to know that I got quite a few of these from Bit-Tech.net. They actually used alot of the same tests, so you can check them out HERE. Just remember that my system is totally different, which is why the scores differ so much at times.
Also, I don't claim this to be a perfect review, but if you have constructive criticism, then I'll see what I can do to make things better.
And I'll let you know now that I have a COD4 test coming up, but I didn't have it finished in time. And I might be able to find some other titles to test as well, but thats not for sure yet.
Once again, I did this because simply, I would have LOVED one of these while I was making the decision to buy mine, and figured that you (and random people searching the web) might feel the same way.
Cheers!
Edited by logan - 3/2/09 at 9:31am
wanting a new motherboard
(13 items) |
|
wanting a new motherboard
(13 items) |
