Memory Bandwidth and Scaling
Everyone should already know that memory bandwidth improves with increases in memory speed and reductions in memory timings. To better understand the behavior of AM2 and Core 2 Duo memory bandwidth we used SiSoft Sandra 2007 Professional to provide a closer look at memory bandwidth scaling.
The most widely reported Sandra score is the Standard or Buffered memory score. This benchmark takes into account the buffering schemes like MMX, SSE, SSE2, SSE3, and other buffering tools that are used to improve memory performance. As you can clearly see in the Buffered result the AM2 on-chip memory controller holds a huge lead in bandwidth over Core 2 Duo. At DDR2-800 the AM2 lead in memory bandwidth is over 40%.
As we have been saying for years, however, the Buffered benchmark does not correlate well with real performance in games on the same computer. For that reason, our memory bandwidth tests have always included an UNBuffered Sandra memory score. The UNBuffered result turns off the buffering schemes, and we have found the results correlate well with real-world performance.
The Intel Core 2 Duo and AMD AM2 behave quite differently in UNBuferred tests. In these results AM2 and Core2 Duo are very close in memory bandwidth - much closer than in Standard tests. Core 2 Duo shows wider bandwidth below DDR2-800, but this will likely change when the AM2 controller matures and supports values below 3 in memory timings as the Core 2 Duo currently supports.
The Sandra memory score is really made up of both read and write operations. By taking a closer look at the Read and Write components we can get a clearer picture of how the two memory controllers operate. Everest from Lavalys provides benchmarking tools that can individually measure Read and Write operations.
The READ results are particularly interesting, since you can see that the READ component of Core 2 Duo performance is much larger than the WRITE results on Core 2 Duo. This is the result of the intelligent read-aheads in memory which Intel has used to lower the apparent latency of memory on the Core 2 Duo platform. Actual READ performance on Core 2 Duo now looks almost the same as AM2 to DDR2-533. AM2 starts pulling away in READ at DDR2-677 and has a slightly steeper increase slope as memory speed increases. The increases in READ speed in Core 2 Duo are a result of the intelligent read-aheads in memory. Performance without this feature would show Core 2 Duo much slower in READ operations than AM2.
This is most clearly illustrated by looking at Everest Write scores. Memory read-ahead does not help when you are writing memory, so core 2 Duo exhibits much lower WRITE performance than AM2 as we would expect. This means if all else were equal (and it isn't) the AM2 would perform much better in Memory Write tasks. Surprisingly the WRITE component of Core 2 Duo appears a straight line just below 5000 MB/s. AM2 starts at 5900 at DDR2-400 and WRITE rises to around 8000MB/s at DDR2-667. Write then appears to level off, with higher memory speeds having little to no impact on AM2 WRITE performance.
Everyone should already know that memory bandwidth improves with increases in memory speed and reductions in memory timings. To better understand the behavior of AM2 and Core 2 Duo memory bandwidth we used SiSoft Sandra 2007 Professional to provide a closer look at memory bandwidth scaling.
The most widely reported Sandra score is the Standard or Buffered memory score. This benchmark takes into account the buffering schemes like MMX, SSE, SSE2, SSE3, and other buffering tools that are used to improve memory performance. As you can clearly see in the Buffered result the AM2 on-chip memory controller holds a huge lead in bandwidth over Core 2 Duo. At DDR2-800 the AM2 lead in memory bandwidth is over 40%.
As we have been saying for years, however, the Buffered benchmark does not correlate well with real performance in games on the same computer. For that reason, our memory bandwidth tests have always included an UNBuffered Sandra memory score. The UNBuffered result turns off the buffering schemes, and we have found the results correlate well with real-world performance.
The Intel Core 2 Duo and AMD AM2 behave quite differently in UNBuferred tests. In these results AM2 and Core2 Duo are very close in memory bandwidth - much closer than in Standard tests. Core 2 Duo shows wider bandwidth below DDR2-800, but this will likely change when the AM2 controller matures and supports values below 3 in memory timings as the Core 2 Duo currently supports.
The Sandra memory score is really made up of both read and write operations. By taking a closer look at the Read and Write components we can get a clearer picture of how the two memory controllers operate. Everest from Lavalys provides benchmarking tools that can individually measure Read and Write operations.
The READ results are particularly interesting, since you can see that the READ component of Core 2 Duo performance is much larger than the WRITE results on Core 2 Duo. This is the result of the intelligent read-aheads in memory which Intel has used to lower the apparent latency of memory on the Core 2 Duo platform. Actual READ performance on Core 2 Duo now looks almost the same as AM2 to DDR2-533. AM2 starts pulling away in READ at DDR2-677 and has a slightly steeper increase slope as memory speed increases. The increases in READ speed in Core 2 Duo are a result of the intelligent read-aheads in memory. Performance without this feature would show Core 2 Duo much slower in READ operations than AM2.
This is most clearly illustrated by looking at Everest Write scores. Memory read-ahead does not help when you are writing memory, so core 2 Duo exhibits much lower WRITE performance than AM2 as we would expect. This means if all else were equal (and it isn't) the AM2 would perform much better in Memory Write tasks. Surprisingly the WRITE component of Core 2 Duo appears a straight line just below 5000 MB/s. AM2 starts at 5900 at DDR2-400 and WRITE rises to around 8000MB/s at DDR2-667. Write then appears to level off, with higher memory speeds having little to no impact on AM2 WRITE performance.
Facebook
Twitter
RSS
118 Comments
View All Comments
OcHungry - Wednesday, July 26, 2006 - link
As far as expensive DDR2 is concerned, apparently you dont understand my logic.You dont need DDR2 800, 1000, and so forth for AMD.
You can buy DDR2 667 which is cheaper and has better timings than DDR2 1000.
With IMC in play, you can and should run FX62 @ 333mhz (FSB and mem speed) using 1:1 ratio. AMD gives us this option thru. IMC and we are not overclocking and being unfair to E6800, just increasing the mem/cpu link to what it should be. Althoug, Running FX @ 200 HTT/FSB compared to E6800 @ 266 would be unfair to AMD. If Intel could run @ 333mhz, so be it.
My calculations are correct and were based on the benchmarks provided by the anandtech review (here).
The only thing that need to change is the % gaine from memory timings (which was not mentioned first in the article). scratching off the memory % gain from my calculation, The new figure put FX62 at about 7% slower than E6800.
But I am amazed how people are forgetting that we are comparing old architecture w/ the new and do not consider K8L or 4x4.
4x4 is due shortly- Lets see if "the new king" willingly hands over the crown.
Must have felt good crowning for less than a month (considering retail not available yet, and less than a month 4x4 is due).
zsdersw - Wednesday, July 26, 2006 - link
I'm amazed that you're expecting us to compare what's soon to be out with what's not-as-soon to be out (K8L, 4x4).If you're going to keep shifting the goal posts to suit your AMD fanboy desires, then you're never going to be happy with anything that doesn't tell you what you want to hear.
Gary Key - Tuesday, July 25, 2006 - link
The true comparison of the FX62 against the X6800 is 2.8GHz against 2.93GHz as they are shipped from AMD and Intel. That is the absolute base comparison and is about as fair, accurate, and truthful as it can get at this time. If you insist on optimizing (overclocking) the FX62 platform then you must realize it is equally fair to do this on the Core 2 Duo platform as well.
The timings used at DDR2-800 were 3-3-3-9. You will not see a 7% to 10% increase in performance.
If the GPU is truly the deciding "facto" for gaming then your logic is up for debate. So, based upon your stated logic it should not matter what CPU is utilized, which equates to Netburst being equal to Athlon 64 in this performance sector. That means Intel was correct for marketing Netburst to the gaming enthusiats. We all know that is not true but are we to now believe it is based upon your statements? It sounds like a $42 Celeron D will perform equally to a $999 FX62 in gaming based upon your logic. I disagree of course but if it is okay to argue that Athlon 64 is a superior performing processor to the P4, then explain to us why is not okay to show Core 2 Duo being a superior processor to the Athlon 64 at this time.
We are performance centric, the same conversations and excitement about a processor series occurred when the FX-51 was released and continued to occur for three years afterwards. Intel has finally countered with an excellent processor series and now it appears to be an issue to discuss peformance advantages without resorting to "My calculations show an overclocked CPU can beat a stock clocked CPU." If you know for a fact that at 9x330HTT and 3-3-3-8 1T settings the FX-62 will match or exceed the performance of a X6800, then please post the screenshots as I am sure all of us would be intrigued by this. However, also post the comparison of your X6800 running on stock voltages at 3.46GHz in order to reflect the true capability of the X6800 also.
I look at it this way, everyone but AMD and Intel win this summer. I know a couple of AMD machines in the house are getting 3800+ X2 upgrades and my primary machine will be sporting a E6600. However, this price matching and slashing is going to have a long term effect on both companies and in the end that might not be healthy for this industry in a couple of years.
Not seeing this difference in build number 5472 of Vista.
OcHungry - Tuesday, July 25, 2006 - link
I see, the moderators keep Editing the review article. the mem timings were not there when 1st read it. oh wellgoinginstyle - Tuesday, July 25, 2006 - link
Dear OC-Sharikou,"Memory timings were DDR2-400 - 3-2-2-5; 533 - 3-2-2-6; 667 - 3-2-3-7; 800 - 3-3-3-9; 1067 - 4-3-4-11; DDR2-1112 - 5-4-5-14." As posted by Wes earlier and hopefully he will place these in the review shortly to shut you the hell up. Your calculations are already incorrect since you ASSumed the DDR2-800 memory was set at 4-4-4-12 and therefore your calculations of the FX62 being another 7 to 10 percent faster are WRONG. Do you understand yet that calculating instead of testing means you are going to be WRONG. You consistently make false and misleading statements here and elsewhere on the internet. We now understand why Intel fired your lazy calculating ass. You were probably the one who devised their P4 marketing scheme to compete against the Athlon 64. I bet all of your calculations at the time proved higher core speeds and memory bandwidth resulted in higher scores against the AMD processors. It was obvious that calculation was WRONG.
Let's see some benchmarks on your FX62 instead of some asswipe musings about percentage based benchmarking. Answer this simple question, why is AMD shipping their CPUs at a 200HTT base? How do you honestly expect to compare a highly HTT overclocked FX62 against a stock Conroe? If their 2.93 Conroe runs at 3.5GHz on stock voltages then would it not be fair to test the Conroe at this speed and the FX62 at your holy 9x330 setting? By the way, in real application land, 1T over 2T might mean a 1% difference at DDR2800, try it yourself. Oh yeah, I f@$&ing forgot you do not own a FX62 or Conroe so you have to use pencil and paper to make your wet dreams come true.
drarant - Tuesday, July 25, 2006 - link
Think of it this way, X6800 will perform about 10-20% higher than a FX-62 whether its a set of DDR2-667 or DDR2-1067, roughly $150 ram vs $250 ram. Whereas, in your proposal, You would need RAM to run at 330 x 4 = DDR2-1320... even if it does exist in the next 6months, it will have to be run @ 2T and 4-4-4 at the tightest.So even though almost all of your speculation is wrong (i.e. 4-4-4 > 3-3-3 does NOT give a 7-10% boost) and the rest is assumption (i.e. 4-4-4 2T is what was used) you still cannot do what you propose to reach a Stock Intel CPU. (Get over it, I did)
zsdersw - Tuesday, July 25, 2006 - link
Pretty much everything you've said is erroneous.zsdersw - Tuesday, July 25, 2006 - link
That was for OcHungry, not Wesley Fink.Wesley Fink - Tuesday, July 25, 2006 - link
When AMD was the performance leader I was called an AMD fanboi, now when Core 2 Duo is top performer I am called an Intel fanboi. That means things are balanced as they should be. Does it ever occur to some that we are fans of PERFORMANCE - carefully compared and measured performance. We did a more than fair comparison overclocking AM2 to the same speed as Conroe and increasing clock speed to 266 to run an unsupported DDR2-1067 - which should favor AM2 as it appears to improve more with memroy bandwidth than Conroe. AM2 still lost by a wide margin.I could even understand your argument if you compared a 4GHz Core 2 Duo, which I am personally running on air, to the highest overclocked AM2 - or even AM2 to Core 2 Duo at the same overclocks and "bus" speeds. I am left to conclude that you surely know more than AMD since they have reduced prices in-line with performance credible review sites have found on AM2 and Core 2 Duo. If AMD believed yours was a realistic comparison the prices would never have been reduced.
An apples-to-apples comparison is not a mystery. Comparing a super-overclocked and bus increased AM2 to a STOCK Core 2 Duo is fanboi.
duploxxx - Wednesday, July 26, 2006 - link
I understand that there are several comments of fanboy'ism that might not be correct, but from the moment you guys went to the click the button visit the comments/conclusions/headers just became no longer "balanced" for example "a new king has born"... the woodcrest isn't that better then the current opteron sure it will have the highest clock cpu and best performance on that 3.0 (when we even can buy it) but compare apples with apples and put a 2,6 versus a 2,66. and making such a statement knowing that the competitor comes with a new revision just a month after.... (not sure if you posted the review btw... its just global anand) same with conroe, it has the performance crown. like amd had it for 3years. never saw a statement when a new fx arrived like "The empire is kicked again". now just new performance blabla.