In light of the fact that there has been some recent A64 OC'ing discussion, I know for sure that the following will be of interest to some (Time and Liam, at the very least).
The following two articles present excellent raw data. Unfortunately, both authours make some errors in logic (amongst other things), so you will have to ignore some of their casual observations and conclusions. However, taken together, they provide really good insight into OC'ing the A64.
[soap box rant]Its unfortunate that the hardware review scene suffers from not following a peer review model. Had the second authour taken the first authour's results (or possibly others) into consideration, a very good analysis could have been derived. Alas, lacking an orgainized method for literature review (ala Medline/Pubmed style), the hardware review scene will likely remain juvenille forever....much to the advangate of the manufactures.[/soap box rant]
Anyways, this first article is quite short. Some of the discussion is poorly explained (i.e. the memory dividers), so don't even bother trying to figure out what he is trying to say. Save your self some time and don't click on his link to the chart either...unless of course you want to see a bunch of 14yrs take several pages to figure out the memory frequency equation I posted earlier (here). Again, the focus should be on the data and the settings he used (which, thankfully, he did a fairly decent job in detailing).
The more recent and longer article can be found here. First thing to know is this guy completely mixes up some of the nomenclature (despite the fact that he wants to try to set the record straight....to his credit, he is absolutely correct in the fact that there is a lot of confusion and misuse between vendors, reviews, public etc. etc. Unfortunately, he himself falls squarely within that group). Anyways, once you realize that he has used LDT where he should have used HTT, things become a lot clearer. Unfortunately, he is not as openly clear in his labels about some of the settings he used - he just doesn't convey all the info, but it is correct nonetheless. He also neglected to mention whether he was manipulating ram timings. By sheer luck, the very first link in a google search of the ram modules he used resulted in his own review of them. I draw from that review that it is very likely that he maintained steady state memory timings when he employded them in this particular review -- with the implication being that the raw data he extracted is all that much more meaningful. And Last of my brief criticisms is that it would appear, by taking the first OC'ing article into consideration, his limited test suite impacted the validity of one of his analysis - ie. although his reasoning regarding the HT bus sounds logical, his conclusion (based on a limited test suite) falls short of what was observed in the first article (larger test suite).
Good science is hard to come by in the hardware review scene. These guys ran some good tests, and achieved some nice results. They just fell short in their presentation and analyses of the results.
The following two articles present excellent raw data. Unfortunately, both authours make some errors in logic (amongst other things), so you will have to ignore some of their casual observations and conclusions. However, taken together, they provide really good insight into OC'ing the A64.
[soap box rant]Its unfortunate that the hardware review scene suffers from not following a peer review model. Had the second authour taken the first authour's results (or possibly others) into consideration, a very good analysis could have been derived. Alas, lacking an orgainized method for literature review (ala Medline/Pubmed style), the hardware review scene will likely remain juvenille forever....much to the advangate of the manufactures.[/soap box rant]
Anyways, this first article is quite short. Some of the discussion is poorly explained (i.e. the memory dividers), so don't even bother trying to figure out what he is trying to say. Save your self some time and don't click on his link to the chart either...unless of course you want to see a bunch of 14yrs take several pages to figure out the memory frequency equation I posted earlier (here). Again, the focus should be on the data and the settings he used (which, thankfully, he did a fairly decent job in detailing).
The more recent and longer article can be found here. First thing to know is this guy completely mixes up some of the nomenclature (despite the fact that he wants to try to set the record straight....to his credit, he is absolutely correct in the fact that there is a lot of confusion and misuse between vendors, reviews, public etc. etc. Unfortunately, he himself falls squarely within that group). Anyways, once you realize that he has used LDT where he should have used HTT, things become a lot clearer. Unfortunately, he is not as openly clear in his labels about some of the settings he used - he just doesn't convey all the info, but it is correct nonetheless. He also neglected to mention whether he was manipulating ram timings. By sheer luck, the very first link in a google search of the ram modules he used resulted in his own review of them. I draw from that review that it is very likely that he maintained steady state memory timings when he employded them in this particular review -- with the implication being that the raw data he extracted is all that much more meaningful. And Last of my brief criticisms is that it would appear, by taking the first OC'ing article into consideration, his limited test suite impacted the validity of one of his analysis - ie. although his reasoning regarding the HT bus sounds logical, his conclusion (based on a limited test suite) falls short of what was observed in the first article (larger test suite).
Good science is hard to come by in the hardware review scene. These guys ran some good tests, and achieved some nice results. They just fell short in their presentation and analyses of the results.