This discussion was and is about the claim of "double" bandwidth, which in the eyes of many is a synonym for double performance. I may have objected to that claim, but you started the pissing match. I'm simply following through.
You have proven nothing. Your ONE(1) metric is NOT supported by other metrics in relation, nor by other reviews. I have challenged you and you answer that challenge with nonsense. Well done. Are you afraid that you'll retest and discover something new?
I was going to let this go but since you keep attacking my creditability, it now time to put the academic hat on. I know you don't like lengthy responses with critical thinking, but please bare with me here.
First lets review the debate at hand to avoid any confusion between both parties. 1) A disagreement between DDR4 and DDR5 bandwidth. The statement that DDR5 provides double the bandwidth over DDR4. 2) Which program is a valid representation of the question above.
1) I back the press release stating DDR5 provides double the bandwidth in
context. If comparing baseline JEDEC specs of DDR4-2133 to DDR5-4800, the numbers I provided by request of you stands true. That note that these frequencies are on the opposite side of the spectrum. Now if you use DDR4-3600 and compare it to DDR5-6400, it is very close to double. Not quite. Move up to DDR5-6600+ and it is. Therefore I conclude of the two scenarios I provided the statement of double the bandwidth is correct.
2) You have pointed out that AIDA64 isn't a truthful representation of bandwidth, I fully disagree. Without any proof of these claims from you, I will do my best to show why you are wrong. First off a margin of error needs to be accounted for. General acceptable margin of error in reviews is 3% unless specified otherwise. "Wildly inaccurate" for myself would start at 10%. 5-19% is questionable. I did ten READs and wrote the values down and did some calculation. The best result was .00005% off the first value and the worst was 3.6%, leaving the average to be 2% margin of error.
I have witness a wider margin on the same system before. That often comes from unstable memory. I also found that a comparison between system can produce a even greater difference in results. For example on Alder-Lake comparing 3GHz to 5Ghz is 3.2%. However It is not linear in nature. 4GHz is closer to 1% from 5GHz. In either case, when comparing results between different computers, you must account for CPU clocks, Ring Cache, IMC ratio, number of DIMMS timings and timings. These all
may impact the overall results depending on the system. In my findings, Alder-Lake (LGA 1700), ring cache and CAS value has no measurable impact outside the already established 2% margin of error. This is why these provided results from my reviews is valid. Same CPU clocks and motherboard for all DDR5. Same CPU and MB for all DDR4. Lock CPU frequency as well.
This leads to the next part of why AIDA64 may be the best representation of memory in reviews. First, anyone can download it and compare without much hassle. Other programs may be paid up front without a trial. Second, if we look at just the theoretical bandwidth per channel, AIDA64 gives a good representation of a synthetic benchmark for RAW memory bandwidth.
We can calculate the theoretical bandwidth of a single channel with this formula. MT/s * 64 /8. DDR4-3600 is 28800 MB/s or 28.8 GB/s per DIMM. If we compare my AIDA64 results which is 58145 MB/s you can see that technically AIDA64 reported over this maximum allotted bandwidth. However it is sill within 3%. Just under 1% to be precise.
While I do not want to discount Passmark or PCMark fully, I do not have those programs to personally test. However just looking at the Passmark database, it seem the transfer rate is based on a single DIMM. 3DMark, PCMark, etc from UL often has arbitrary scores for easier general comparison between system. Though, you can miniplate scores easily increase that margin enough that it is no longer a good representation of a system. Though in this instance I cannot say for certain with PCMark10.
So there you have it. I provided data when you asked for it, three times now and when I asked repeatedly for any sources from you, I received nothing. By defacto I won this debate a longtime ago, but you keep insisting that my viewpoint is wrong with zero credibility. Now the tables have turned. I request that you provide evidence with explanations to back up your claims. Otherwise, you can walk away from this with the tale between your legs. In the future make sure you can at least properly debate someone before digging your heels in. Please do not spread lies about my creditable anymore. You clearly lack the abilities to make valid arguments.
