Originally Posted by hurricane28
Originally Posted by d1nky
similar to mine ATM but i got 2050mhz ram with 10-10-10-28
and i cant decide between 2550mhz ram and 11-13-13-31 which is slightly better
2600mhz on cpunb 1.35v not worried about a high HT with a single card/1 ssd etc
I set mine at RAM profile 2400 and loosened the timings a bit and set the multipliers the highest i can get.
I did noticed that those chips well at leas mine does not like more than 2600 CPUNB and much over 3000HT.
HT link is not only for GPU's it does lots more that is why its called hyper transport link.
It also is important for RAM speed etc.
here some information about it: http://www.overclock.net/t/644852/hypertransport-link-myth
That link you provided really does not further your point. the "astounding increases" he is on about are all very easily within the margin of error (1.2%, 1.1% 1.2%. .9%.5% and one test he describes as "mind boggling" was 2.6% increase.
add a couple tests that were slower with the 3300HT
this mind you with a extremely OC HT by 60% running at god knows how many volts, and a CPU OC by 37% with 1.65v being pumped through it. I would also be willing to bet that the tests that showed a extremely meager increase were the second run.
I hardly think this is in any way shape or form a 'mythbuster'
A better way to look at the HyperTransport is that it allows data to not be slowed down rather than actively making them faster, and does this by removing system busses and the corresponding latencies. In other words if you have a handful of cars sharing a six lane freeway, adding a seventh lane is not going to make a difference. Before the advent of Hyper-transport you had a lot of cars trying to share a single lane for traffic both ways as it were.
As a reviewer we rerun benchmarks between 3 to 7 times and take the average score because we will very often get differences of this (or more) and while running the same exact settings, with restarts, and a clean OS install with precisely the same number of processes running for each benchmark run. We occasionally will use not the mean, but the median when an anomalous run happens which further demonstrates the importance of methodology.
What we have here is a series of benchmark runs made on a system with a 60% NB OC, 37% CPU OC, on a 'Dirty system that averages out to under a 1% difference. (not to mention with those voltages and the resulting chipset/CPU/VRM temps, who knows what was happening) All this while running intensive benchmarks trifire with three of the biggest baddest cards at the time, he was close to the maximum data that could be put through a system at the time.
Here is some comprehensive information about HyperTransport:http://www.amd.com/us/products/technologies/hypertransport-technology/Pages/hypertransport-technology.aspxhttp://www.hypertransport.org/docs/uploads/HyperShare_Intro.pdfEdited by Red1776 - 7/15/13 at 6:51am