(05-10-2013, 01:27 PM)Venomx1 Wrote: [ -> ]dammmhn that 5.3ghz sandy basically matched my 3570k blow for blow (and runs same speed ram as i do). cept it was faster in first save.
jelly 
.
sadly i think i damaged my ram after uploading those Overclocked memory scores.
111/84/132.
thing isnt stable at 5ghz anymore. mmm
RAM speed doesn't matter for the majority of aplications when it comes to Ivy/Sandy Bridge cpu. Dolphin included, I think. Also, an Ivy Bridge is only around 5% faster than a Sandy Bridge at the same clock in most apps too. That's the reason my results matched yours.
Actually ivy bridge is 10% faster than sandy bridge at the same clock rate in most applications. For dolphin it seems to be closer to 0-2%. Not every application benefits from the improvements made to the microarchitecture in ivy bridge, and it appears that dolphin is one of those applications.
(05-13-2013, 02:42 AM)NaturalViolence Wrote: [ -> ]Actually ivy bridge is 10% faster than sandy bridge at the same clock rate in most applications. For dolphin it seems to be closer to 0-2%. Not every application benefits from the improvements made to the microarchitecture in ivy bridge, and it appears that dolphin is one of those applications.
i think its closer to 6% better per clock overall in applications. (not counting dolphin).
Tomshardware, xbitlabs, and anandtech all say 10% and have the benchmarking data to back it up. So no offense but I'm going to trust them instead of you.
When Gabriel Belmont changed from a 2600K to a 3770K, he said he got a 'big' speedup in dolphin, without overclocking either. Unless it was the placebo effect, I'm going to say Ivy is a it more than the 2% better for dolphin you said, especially as they were by the same user on an otherwise identical rig.
Gabriel's performance could have been impacted by cooling. I noticed a pretty nice boost on my 1100t by putting on a good cooler. Also I think turbo may work better on Ivy giving it a bit of an advantage on non overclocked systems. For instance NV was getting +300 MHz turbo on his Ivy running dolphin while Shonumi and I where only able to get +100 MHz on our Sandys. So 3.7 on NV's 3570k and 3.4 on our 2500ks. An advantage of 300Mhz for the Ivy. Same clocked OCd systems wouldn't have that advantage and this benchmark shows they perform almost identically.
I'm not saying this benchmark should be taken as gospel; due to all the variables.
They were both water cooled, so should have both got to maximum turbo.
Except Sandy Turbos thusly:
Using a 2500K as example
3.3GHz stock
3.4GHz 4 cores
3.5Ghz 3 cores
3.6Ghz 2 cores
3.7Ghz 1 core
A 2600k is similar but +100Mhz on each of the above stages.
It is impossible to hit max turbo regardless of cooling unless only one core is being used. Both Shonumi and I got the same 3.4GHz turbo boost while using dolphin, though mine got very infrequent short bursts to 3.5.
Cooling is absolutly not an issue on my pc. OCd to 4.5Ghz my cores only hit around 54c while stress testing. Real world use has never seen it hit 50c.
Contrast this with NV's 3570k which ran at 3.7Ghz while running dolphin. It would seem that Ivy turbos a bit better. So his 2600k was probably boosting to 3.5 while the 3770k probably boosted to 3.8.
By max turbo, I thought it was implied that he was hitting max turbo for the number of cores being used, which theoretically would be the same.
For me max turbo means the maximum turbo frequency the cpu will use, but I see how you mean it now. Still they would not be theoretically the same.
First of all the Ivy counterparts have a 100Mhz advantage to begin with. Then the turbo levels are different too.
2600K
3.4Ghz stock
3.5Ghz on 4 cores
3.6Ghz on 3 cores
3.7Ghz on 2 cores
3.8Ghz on 1 core
3770k
3.5Ghz stock
3.7Ghz on 4 cores
3.8Ghz on 3 cores
3.9Ghz on 1 or 2 Cores
Even if they both boosted to the 3 cores level Ivy would have a 200Mhz lead over Sandy.
Now take into account that NV's Ivy boosted to 3.7Ghz in dolphin, the 3 cores level for his cpu, and me and shonumi's Sandy only boosted to the 4 core level. Seems to suggest that Ivy boosts a little better in dolphin than Sandy does. Granted this is a small sampling and really can't be used to confirm or refute anything.
My point in all this is that you are saying NV's estimate of ~2% performance increase is wrong based on the comparison of two chips with dissimilar clocks. Of course Gabriel noticed a performance increase he was running a chip that was at least 200Mhz faster in addition to being ~2% faster per clock.