(11-16-2011, 04:30 PM)NaturalViolence Wrote: [ -> ]First I need to make sure you understand what the clock signal is and how it's used, I think I did a fairly good job explaining this in another thread: http://forums.dolphin-emu.org/showthread.php?tid=19672&page=4
Post #37, read it then post when you're finished so I can explain further.
Fascinating.
So does i5 make each oscillation do more?
This is fairly confusing.
(Feel free to post, I gotta go to bed.)
Quote:Hey, why is i5 2300 good for emulation despite not crossing the 3 ghz threshold?
It does actually cross the 3GHz threshold thanks to turbo boost. It turbo boosts to 3.1GHz and can turbo boost to 3.4GHz with a limited turbo overclock (which the motherboard you picked supports).
Quote:So does i5 make each oscillation do more?
No, at least not in the sense that you're probably thinking of.
Quote:What makes sandy bridge so efficient per clock or whatever?
Better overall microarchitecture. Microprocessors are a very complex chip that attempts to integrate all of a computers processing needs into a single set of programmable circuits. Modern microprocessors contain literally billions of circuits to do all kinds of things, they are far to complex to be truly well understood by the likes of you or me. Since there are so many variables to examine when comparing different microprocessors (many of which we don't even know about since a lot of the details of the microarchitecture have to be kept secret so nobody can copy them) the best way to compare two chips is to run a bunch of different tasks on both of them and see which one does it faster.
The simplest way to analyze that is
(number of instructions to do the task * average number of clock cycles needed per instruction) / number of clock cycles per second = total time needed to do the task
Of course just because a particular microarchitecture is better at one task doesn't mean it will be better at another. And since we're comparing only x86 cpus here (therefore they all use the same machine code) we can say the number of instructions to do the task is always the same. Therefore it falls to the clock rate and the number of clock cycles needed to complete each instruction to determine performance.
Sandy bridge is just a better microarchitecture with better circuit logic (the design of circuits needed to accomplish a particular task) and thus can perform certain actions in fewer clock cycles. I could list a lot of advantages but I'm tired now and I think I already did that earlier in this thread.
Quote:Isn't 2.8 ghz bad for emulation?
Well no. An N64 emulator certainly doesn't need a 2.8GHz cpu. But if you mean to say dolphin, yes, but it depends on the specific cpu microarchitecture, how demanding a particular game is, and what settings you're using.
You did indeed post the advantages of sandy bridge earlier on.
My bad.
"Just some of the many many variables to look at
External Bus Speed: sandy bridge wins
Memory speed: sandy bridge wins
Pipeline width: sandy bridge wins
Branch prediction effectiveness: sandy bridge wins
Cache size: deneb (phenom II) wins
Cache bandwidth/latency: sandy bridge wins
Clock rate: deneb wins (in this circumstance)
Number of cores: equal
Maximum integer throughput: probably about equal thanks to the higher clock rate
Maximum floating point throughput: sandy bridge wins
Maximum SIMD throughput: sandy bridge wins
Functional unit utilization (percent idle time): sandy bridge wins"
Just so I'm clear:
An AMD Phenom has a 1 ghz deviation, right? (3.8 on AMD = 2.8 on intel.)
From the PCSX2 website.
http://forums.pcsx2.net/Thread-CPU-Benchmark-designed-for-PCSX2-based-on-FFX-2
59.26 FPS - SLUS 20672 - AMD Phenom II X4 955 - 3.96 GHz OC - Ryner Lute
60.84 FPS - SLUS 20672 - Intel Core i7 2720QM - 2.2 GHz Stock (3.0 GHz w/TB) - Hippolytus - CPU-Z (laptop, but equal to mine.)
See what I mean?
How far or close is the gap for bulldozer?
Ugh, to many things to point out, I'm too tired to do it all now. I'll post as much as I can before I log off.
1. I don't think you fully understand the word deviation. The word deviation does not fit with that rest of that sentence.
2. Differences should be shown as a percentage. 1GHz and 2GHz is a 1GHz difference and a 100% difference. 1.5GHz and 2.5GHz is a 1GHz difference and a 66% difference. 2GHz and 3GHz is a 1GHz difference and a 50% difference. 2.5GHz and 3.5GHz is a 1GHz difference and a 40% difference. 3GHz and 4GHz is a 1GHz difference and a 33% difference. See how those are all 1GHz apart yet they are all completely different ratios?
3. This assumes no external bottlenecks and a constant linear relationship between core clock rate and performance. Which is impossible. Just because a 3GHz and 4GHz cpu get the same performance that does not mean the 3GHz cpu will always have 1/3 higher performance per clock at any clock rate.
4. You're using one specific test for a comparison. This is nowhere near enough to claim the superiority of one microarchitecture over another.
5. The results of the test cannot be verified (although most of them are probably accurate) and minimal details are given by users. If you attempted to cite this benchmark as empirical evidence in a thesis I would laugh.
6. That test is optimized for newer SSE extensions available only on intel cpus and makes extensive use of them, most software doesn't even have a use for these instructions.
Most of those results are what I would expect. It does not surprise me that a sandy bridge cpu at 2.2GHz is keeping up with a phenom II 955 clocked at 3.98GHz in that particular test. This test is literally perfectly optimized for taking advantage of sandy bridge's wide pipeline and excellent SSE performance. You could not have designed a better test to show off the advantages of the sandy bridge microarchitecture if you tried.
Quote:How far or close is the gap for bulldozer?
Bulldozer has less single/dual threaded performance than phenom II. It only beats phenom II in heavily multithreaded software (just like when phenom was released, and most software is not heavily multithreaded as I'm sure you're aware). However it does support the newer SSE extensions so that might even things out a bit in that particular test. I can tell you it would definitely perform worse than phenom II in dolphin.
Now this question is directed at the entirety of the Dolphin community.
Which of these is more likely:
(When I use the i5 2300, I'll find that more games run full speed than those that don't)
or
(When I use the i5 2300, I'll struggle to get anything to run at full speed.)
Sorry, even with all of this science-y stuff, I'm a bit of a pessimist.
As a side note- how do you overclock a NON-K Sandy Bridge i5?
Alrighty.
Today's buying day.
Here's what I have:
GTS 450
I5 2300 + ASUS P8H67-M PRO/CSM
G.SKILL Ripjaws Series 8GB (2 x 4GB) RAM
Antec BP550 Plus 550W Continuous Power ATX12V V2.2 80 PLUS Certified Modular Active PFC Power Supply
Tuniq TX-2 Cooling Thermal Compound
This is to be bought in a few hours.
One last question....
is
this viable?
You should be fine with the 2300.
You can overclock the NON-K versions, but only by a few MHZ, like a hunderd or so.
The reason why, is because its locked.
If you plan to overclock, get a K version.
I'm sorry, I haven't read the whole thread.
But what do you mean with "is this viable?".
Do you want to SLI or something?
The rest of the rig looks good.