I have loads of examples from my Benching to back this up. They are countless. I just dont have time to bench inefficient systems and then efficient ones to back up this stuff. This Cinebench example I am presuming is all none tuned XMP. Tuning thirds alone can make a large difference to scores. Here they are looking at frequency and primary timings not anything else = rubbish. How about DDR2133 C6 vs 1600C6 and then see.
8 pack upgrade question
- Thread starter clayer
- Start date
Gentleman, I specifically asked 8pack about video and how memory timings affect that. Specifically 1600Mhz Vs 2133Mhz. His reply was that Cinebench (which has a 3D video benching component) was 'loads faster'. I will say it straight; 8packs statement that
All the research I've done and all my testing shows it is NOT loads faster in Cinebench and similar for other video applications like MeGUI, AVIDemux etc. I work with video and the simple truth of the matter is that in the video framework there is no point getting faster memory than 1600Mhz or tightening up the timings more than usual default 9-9-11-24 for a more 'efficient' system as it will not make much difference to transcoding/encoding speeds.
From Anandtech.
If you are a SuperPI bencher etc then by all means go to town and get as fast as memory as you can and lower the latency's as you will definitely see improvements. Though if you're using SB for video/gaming spend the money elsewhere.
is false and incorrect. No big deal, I've been wrong before and will be again. Which is why I asked for evidence, unfortunately there was non forthcoming.
All the research I've done and all my testing shows it is NOT loads faster in Cinebench and similar for other video applications like MeGUI, AVIDemux etc. I work with video and the simple truth of the matter is that in the video framework there is no point getting faster memory than 1600Mhz or tightening up the timings more than usual default 9-9-11-24 for a more 'efficient' system as it will not make much difference to transcoding/encoding speeds.
From Anandtech.
If you are a SuperPI bencher etc then by all means go to town and get as fast as memory as you can and lower the latency's as you will definitely see improvements. Though if you're using SB for video/gaming spend the money elsewhere.
Last edited:
LOL your still missing the point man!! I said when I tested Cinebench 2133mhz was loads faster and it was. I tuned the sticks!!!
Do you think I am clueless about memory timings and efficiency like the above example thats been given. The 1600mhz top example is PSC IC which is not your average 1600mhz kit its high quality. Average Kit is 9-9-9-28 2T. PSC can beat 9-9-9-28 easy at 2600mhz!!! This slide has not shown that. Its shown hynix at 2133 which cant do the tight timings so therefore your not comparing memory speeds are you. Your comparing speeds and timings. The reviewer if you want to call him that should have shown the same timings at all speeds to show how memory speed effects performance. Instead quality IC at 1600mhz= pointless.
7-11-7-28 1T at 1666mhz is far inferior on IB CPU that the same timings at 2133mhz on the exact same sticks.
Do you think I am clueless about memory timings and efficiency like the above example thats been given. The 1600mhz top example is PSC IC which is not your average 1600mhz kit its high quality. Average Kit is 9-9-9-28 2T. PSC can beat 9-9-9-28 easy at 2600mhz!!! This slide has not shown that. Its shown hynix at 2133 which cant do the tight timings so therefore your not comparing memory speeds are you. Your comparing speeds and timings. The reviewer if you want to call him that should have shown the same timings at all speeds to show how memory speed effects performance. Instead quality IC at 1600mhz= pointless.
7-11-7-28 1T at 1666mhz is far inferior on IB CPU that the same timings at 2133mhz on the exact same sticks.
How on earth can I be missing the point. 
I've asked the question
Which most people would take to mean 2133mhz is loads faster than 1600mhz using the same timings because that is what my question was.
Let's surmise that by the term 'loads' most people would understand it to be >20% faster... so let's be generous and come down to 'loads' meaning >10% faster.
It's so easy to prove or disprove (this is why I preferred studying Maths over English) no need to throw out lots of figures to obfuscate the central point. It will take 5 minutes. Run Cinebench with memory at 1600Mhz and 2133Mhz with the same timings and tell us what scores you get.
I've asked the question
Part of your reply...
Which most people would take to mean 2133mhz is loads faster than 1600mhz using the same timings because that is what my question was.
Let's surmise that by the term 'loads' most people would understand it to be >20% faster... so let's be generous and come down to 'loads' meaning >10% faster.
It's so easy to prove or disprove (this is why I preferred studying Maths over English) no need to throw out lots of figures to obfuscate the central point. It will take 5 minutes. Run Cinebench with memory at 1600Mhz and 2133Mhz with the same timings and tell us what scores you get.
8pack you do sterling work on this forum and give great advice to many people.
It is a tad bit disappointing to see that you may be suffering from the hubris that sometimes afflicts people with influence. That sense of infallibility, that even when you make a small mistake you can't be seen to be wrong. (The recently departed Margaret Thatcher comes to mind...)
Keep up the great work and remember, the measure of man is not in the size of his traps.
It is a tad bit disappointing to see that you may be suffering from the hubris that sometimes afflicts people with influence. That sense of infallibility, that even when you make a small mistake you can't be seen to be wrong. (The recently departed Margaret Thatcher comes to mind...)
Keep up the great work and remember, the measure of man is not in the size of his traps.
It actually is and his arms.
There is no mistake many moons ago I did a memory thread showing this. Cinebench does not interest me even remotely as its not on the bot for points.
I will still run it though 5ghz 1600mhz RAM vs 5ghz 2600 mhz 7-11-7 RAM for sumat that remotely interests me to do!!!!
I am busy with Haswell and developing boards and bios at the mo so may even do it on Haswell to make it more meaningful for users of the new platform.
There is no mistake many moons ago I did a memory thread showing this. Cinebench does not interest me even remotely as its not on the bot for points.
I will still run it though 5ghz 1600mhz RAM vs 5ghz 2600 mhz 7-11-7 RAM for sumat that remotely interests me to do!!!!
I am busy with Haswell and developing boards and bios at the mo so may even do it on Haswell to make it more meaningful for users of the new platform.