New Posts  All Forums:

Posts by xlink

1 week after I buy a new card... of course... I still paid less than their current MSRP (even after rebate) though because I shopped around.
@DargonplayTimings don't matter in and of themselves, it's timings/clock speed.20/4000Mhz = 10/2000Mhz = 5/1000Mhz = 2.5/500MhzWe've actually gotten slightly BETTER on timings RATIOS each generation.
I think he assumed no AC.And yes, 110F would help cool a 160F CPU... and possibly fry you in the process.
If you're comparing against 2005/2006... it's true that 10 years passed (roughly 5 2 year periods)but that isn't all there is to CPU design. Each iteration of CPU has more cache(which takes up space) than its predecessor and there are additional constraints to handle.The transistor count has more or less kept pace. It hasn't translated to MOAR COARS.
actually... on servers it matters in some cases.Suppose you have a 100 core/thread server and have a task broken into 100 pieces which must be completed before going onto the 2nd task.Your total completion time will be equal the MAX of those 100 pieces.Assume each task is ~ iid Normal, mean 5s sd 1...If you ran each on an identical core/memory subsystem, you'd AVERAGE around 7.5s total completion time. If you could turn off cores as they completed and instead turbo up 50%,...
It's almost like making huge dies is expensive and not worth the tradeoffs for most players... I think Intel has different dynamics from IBM(such as superior fabs and higher volumes) and can thus afford to focus have segments with higher die sizes.
I'd be happy with the same game but the rendering engine shifted up to 8x the resolution with all of the sprites being scaled as well.
Can't wait to get 10 years into the future. There's been a lot of things promised that can't come soon enough. Memristors, super-effective batteries, etc.
PCI-e is used to communicate between components in the system. For heavier loads, less time is spent communicating and more time is spent processing the data that needs to be sent out. In a sense, PCI-e bottlenecking occurs when it matters the least - when workloads on the GPU are low.It might actually be a minor issue. SLI/CF require a lot of communication over the bus. The memory in both cards needs to match up and that data is sent over PCI-e.
I need the double x16 PCI 3.0 for my quadruple SLI-CF though that I plan on getting in 5 years (currently broke, spent too much on the motherboard).
New Posts  All Forums: