New Posts  All Forums:

Posts by TranquilTempest

The first consideration for CPU is pure perf/$, but once you have a basic outline you look at your bottleneck and consider alternate parts as a proportion of total system cost. Most of the time that means picking a 4 core chip with the best single thread performance the budget allows.
like techpowerup?
It's not an issue of real vs ideal, it's a question of whether the test was properly controlled, or confounded by two IHS that were different shapes. The IHS on your 4930k might be shaped like his 2600k, maybe it's not just about big die vs small die. If you don't have controlled tests measuring the impact of each different variable, you don't find out what actually causes the difference. The point is, if you're comparing two different coolers, you use a heat source you...
Now you're talking about the interface between the cooler and the IHS, not the performance of the cooler. Were both CPUs lapped flat and used with the minimum possible quantity of thermal paste?Interface conductance is the paste thermal conductivity * surface area / thickness. This means halving the thickness of the paste layer is equivalent to doubling the thermal conductivity of the paste. Basically, flat is always better than bowed unless you're using too much...
Any information on what exactly kaby lake will have over skylake?
That's not how it works. You can't test performance without measuring how much energy is dissipated through the cooler. The biggest 'problem' with anandtech's test method(compared to a "real world" method) is that their heat spreader may be more effective than the one on a typical CPU, so there may be a slight edge for coolers that don't concentrate their cooling paths directly over the die, but I don't think that issue is anywhere close to the variability of thermal...
Noise is a grab bag with non-reference cards, though it does tend to correlate with power usage. 380x is better than the 280x in that regard, though still nowhere close to fury or maxwell in terms of perf/watt.I seriously doubt the vram quantity is why there's a perfomance difference in some games, the architectural differences should matter much more.
Just buy some earplugs?
There's a good chance that 3 minutes isn't relevant, the operator could deburr this part while the next one is in the machine.
HT is 10%~25% improvement in fully saturated performance, doubling the number of cores is a ~100% improvement(provided you have the same clock rate). If I'm paying 50% more for a processor, I want more than just HT out of it.
New Posts  All Forums: