I am fairly tech savy, knowing percentages differences between cpus, such as ivy bridge being around 6% faster than sandy bridge, but having a max overclock of around 4% lower, and using around 20% less power on the max overclock. i know of skylake, bringing mainstream quad core and so on. I read the latest rumours on upcoming new cpus, such as haswell refresh on the new intel 9 chipset in april/may which will support broadwell with a bios update.
The games i play are single threaded bottlenecked. I play Oblivion heavily modded (uses 1.3 threads), X3 albion modded, (uses 1.1 threads), Anno 1404 (uses 2.5 threads ), Earth 2150 with custom camera angles (1 thread), Crysis 1 modded (2.2 threads). I have a cpu and gpu monitoring program at all times.
I have tested the bottleneck by overclocking and downclocking the gpu and cpu. I run usage monitoring 24/7 for gpu and cpu for each individual thread including cross references to volatage used by cpu and gpu, and total system power draw. i know exactly when the system is bottlenecked and what is the cause, cpu or gpu.
i was getting in those games as low as 20 fps on my overclocked nehalem i7 920, which i replaced for an i3220, long story, i quit gaming for a while. The fps i got from a 4 ghz overclocked i7 920 is around the same as the i3220 at 3.3 ghz due to better ipc, and because the games i play use at most a little over 2 threads.
When i have 20 fps my gpu is at something like 40% utilization, and overclocking the gpu and turning AA off makes no difference, just drops gpu usage even lower, to 25%. The cpu is hogged with scripts it has to run, and distant land geometry and high number of polygons on an engine that is not made to render so many.
I have read all discussions about haswell i5 vs i7, i am not asking for the question in that sense. i know that the extra 2mb cache the i7 has vs an i5 gives around 2% performance in gaming at same frequency. i have analyzed more benchmarks with a calculator. However, i did not see extensive comparison between the max overclock of an i7 vs i5 and i see none with ht and disabled cores. Here is my question:
i would probably get a haswell refresh cpu k series that will get released on 2H 2014. Based on the current haswell cpus, which one overclocks better after disabling HT on air, the i7 or i5? is this the best single threaded performance possible? i will disable cores as well. yes, i would get the i7, disable ht and 2 of the cores, overclock it as far as i can and play oblivion. There is no other way but to brute force it on 1-2 threads. i would get a custom air cooler.
If i7 is a better overclocker after disabling 2 cores and ht, i would buy it over the i5. If it is a 2% improvement with the extra cache and a 4.6 vs 4.4 max overclock on 2 cores, that is a 6.6% more fps. Higher cpu frequency translates into direct fps improvements when heavily cpu bottlenecked, i have done benchamarks in each of those games with various gpu and cpu oc/underclock.
Also what is your prediction of haswell refresh improvement? Something like a 5% higher max overclock at same voltage and temperature vs current haswell k models ? Speaking of future models, broadwell will have around 30% less power consumption, but will it transalte to any performance increase? My guess is that it wont due to intel focusing on low power mobile and 14nm process being immature.
tl;dr Max single threaded performance possible on air on 2 cores, combination of ipc and high frequency. I am looking for extensive testing with the cpu overclocked on air and single threaded results in multiple benchmarks. HT has to be disabled, and best if part of the cores are turned off for better overclocking.
The games i play are single threaded bottlenecked. I play Oblivion heavily modded (uses 1.3 threads), X3 albion modded, (uses 1.1 threads), Anno 1404 (uses 2.5 threads ), Earth 2150 with custom camera angles (1 thread), Crysis 1 modded (2.2 threads). I have a cpu and gpu monitoring program at all times.
I have tested the bottleneck by overclocking and downclocking the gpu and cpu. I run usage monitoring 24/7 for gpu and cpu for each individual thread including cross references to volatage used by cpu and gpu, and total system power draw. i know exactly when the system is bottlenecked and what is the cause, cpu or gpu.
i was getting in those games as low as 20 fps on my overclocked nehalem i7 920, which i replaced for an i3220, long story, i quit gaming for a while. The fps i got from a 4 ghz overclocked i7 920 is around the same as the i3220 at 3.3 ghz due to better ipc, and because the games i play use at most a little over 2 threads.
When i have 20 fps my gpu is at something like 40% utilization, and overclocking the gpu and turning AA off makes no difference, just drops gpu usage even lower, to 25%. The cpu is hogged with scripts it has to run, and distant land geometry and high number of polygons on an engine that is not made to render so many.
I have read all discussions about haswell i5 vs i7, i am not asking for the question in that sense. i know that the extra 2mb cache the i7 has vs an i5 gives around 2% performance in gaming at same frequency. i have analyzed more benchmarks with a calculator. However, i did not see extensive comparison between the max overclock of an i7 vs i5 and i see none with ht and disabled cores. Here is my question:
i would probably get a haswell refresh cpu k series that will get released on 2H 2014. Based on the current haswell cpus, which one overclocks better after disabling HT on air, the i7 or i5? is this the best single threaded performance possible? i will disable cores as well. yes, i would get the i7, disable ht and 2 of the cores, overclock it as far as i can and play oblivion. There is no other way but to brute force it on 1-2 threads. i would get a custom air cooler.
If i7 is a better overclocker after disabling 2 cores and ht, i would buy it over the i5. If it is a 2% improvement with the extra cache and a 4.6 vs 4.4 max overclock on 2 cores, that is a 6.6% more fps. Higher cpu frequency translates into direct fps improvements when heavily cpu bottlenecked, i have done benchamarks in each of those games with various gpu and cpu oc/underclock.
Also what is your prediction of haswell refresh improvement? Something like a 5% higher max overclock at same voltage and temperature vs current haswell k models ? Speaking of future models, broadwell will have around 30% less power consumption, but will it transalte to any performance increase? My guess is that it wont due to intel focusing on low power mobile and 14nm process being immature.
tl;dr Max single threaded performance possible on air on 2 cores, combination of ipc and high frequency. I am looking for extensive testing with the cpu overclocked on air and single threaded results in multiple benchmarks. HT has to be disabled, and best if part of the cores are turned off for better overclocking.
Last edited:
