Dynasty Warriors 3 speed issue

[tuanming]

AMD still hasn't quite fully beat the Core 2 Duos and Quads because they're beating a dead horse on that old arse architecture. Generation wise, I'd say AMD is 3 generations behind in terms of performance.

[nosisab Ken Keleh]

I forgot to say the I5 are the ones competing with the Phenom II cost/performance wise. No doubts the Intel top line puts the phenom to shame, a shame their prices put me to shame...

I don't know how it is in US but here those prices create a monstrous abyss.

And this difference in prices is not limited to the CPU themselves, all the remaining hardware is expensive on Intel platform too.

Besides, if the tendency is not stopped, soon we'll have PC with the same "hardware protection" found in Sony products... thank you so much Intel, you and TCG.

[Aastii]

It is a fact. Clock for clock intel beats amd, and it has been this way for several generations. Amd only beats some intel cpus when they are clocked higher, or have more cores. I have both intel and amd cpus and know this from experiance. If you want a cheap pc amd is great since it offer a lot of choices in the low price ranges, but if you want maximum performance you have to go intel.

Sorry to agree here, but yeah, AMD chips have consistently be lots slower than Intel's for the last couple of years.

You are trying to tell me a Phenom II is slower than a P4 or Celeron?

As I said, some Intel chips are faster than AMD and the other way around. I agree that for the moment, Intel holds the title of fastest mainstream CPU on the market, that doesn't mean, as you are trying to say, ALL current AMD chips are slower than all Intel chips.

Gamergeek, I currently have several systems with both AMD and Intel in, and also know from experience that not all Intel chips are faster than all AMD. I don't favour either, both AMD and Intel make excellent chips, all of my systems, be it my Athlon64 x2, Phenom II, Core2 or my i7 builds all perform excellently for what they are designed for, but out of them, the AMD chips are by no means slouches, and are by no means the slowest of my builds.

I think though, come June with Bulldozer taking over from K10, Intel could have a struggle on keeping the top position if AMD live up to expectations, and you could see ourselves going back to AMD dominance of all areas

[refraction]

You are trying to tell me a Phenom II is slower than a P4 or Celeron?

did you seriously just compare a modern chip to a redundant no-longer made chip to make a point?

anyway, the point he was making, the clock for clock is slightly better on intels these days, however, in terms of the emulator, we take advantage of SSSE3 and SSE4, both of which AMD's don't support (they have SSE3 and SSE4a, which are useless pretty much), which in our case gives intels a much higher performance advantage than AMD's can get.

[Saiki]

did we all go completely off the topic here?

well, I'll drag iut back on topic

I tested DW5 on my PC and I get a 100% EE at 25 fps, but when I rebuild and activated a different kind of hack (not a speedhack, a slowdown hack, don't ask. the devs know what I'm talking about) I hit on average 30fps, and my EE was at about 60% or so

[rama]

The topic is a good one, for I can add the One Big advantage Intel chips have *in PCSX2*.
Remember that we rely on SSE heavily, especially with the demanding VU recompiled code.
There's a small and relatively unknown optimization called "Denormals are Zero" built into AMD and Intel CPUs
that speeds up processing of otherwise computationally expensive "denormal" floating point values.
PS2 games regularly run into these denormals.

So the big thing with AMD here is:
Compared to Intel, their "DaZ" implementation is just barely faster than not doing "DaZ" at all.
We can only speculate why that is, but it effectively gives Intel an advantage of 5 to 30% over AMD.

So yea, even if AMD manages to compare to Intel in some popular benchmarks or PC games, it still has
that large performance problem in PCSX2 to make up.
(I really hope Bulldozer will fix this, but we'll see )

[Aastii]

did you seriously just compare a modern chip to a redundant no-longer made chip to make a point?

anyway, the point he was making, the clock for clock is slightly better on intels these days, however, in terms of the emulator, we take advantage of SSSE3 and SSE4, both of which AMD's don't support (they have SSE3 and SSE4a, which are useless pretty much), which in our case gives intels a much higher performance advantage than AMD's can get.

P4's aren't made, but Celeron's still are, but either way that is my point, both are still frequently used chips, 1 is still a chip that is mass produced, yet neither gets close to standing up to some AMD chips, so it straight away disproves the fanboys saying "Intel > AMD...fact!!"

So far as in PSX, the reasons behind Intel being the better option is very interesting and thankyou to yourself and rama for pointing those out.

In every other game I have ever tried I can hit constant 60 fps (NTFS games) just seems Dynasty Warriors doesn't want to play unless I throw a higher, but less stable OC on there. Temperatures aren't an issue, I'm only at mid 40's, but voltages to get it stable at 3.7+ GHz is, you are talking going 1.5V+, which for a Phenom II, it is risking frying the chip. I've killed a couple on lower voltages with temperatures similar to this one .

Aah well, Will have to go and find my PS2 composite cable and dig out the console itself to play this one from now on

[nosisab Ken Keleh]

Now comes the ironic part. No doubts about "today" CPUs favoring Intel on PCSX2, mainly because the SSE thing...

But words are the incoming Bulldozer will change this, the reason being first and foremost, they will have all that instruction sets and more yet. But it's not all, the bulldozer architecture provides two 128 bits float points registers shared by core (which actually are two complete independent integer cores) which can be used separated or as one bigger 256 bits register against the 256 bits registers from Intel.

The point is most used instructions use only those 128 bits which benefits from having two discrete 128 bits registers available, besides the same instructions perform worsen in 256 bits registers. Time will say but chances are the new AMD architecture may prove a better deal for the nowadays PCSX2 implementation.

Bulldozer design clearly suggest AMD is wishing to not compromise the CPU with features better fit on the GPU and a signal they are going to implement the two concepts together and preparing for it.

[tuanming]

The differences between SSE2, SSSE3, and SSE4.1 are very minimal to none. This was tested with FFX-2 NTSC, notice how I didn't used a demanding game. Well, it was for the purpose of a more accurate data reading Seriously, what games offer substantial or significant performance gain when using SSSE3 and SSE4.1?

[nosisab Ken Keleh]

Although I never tested it myself (mainly because I don't own an Intel machine from the last 2 decades at least). It's supposed to give a peak at max of 10% in certain games... more yet, the amount of performance gain varies across the same game... it's not a fixed value.

Still it could mean 4 to 6 frames more to FPS (in the best scenery) where the machine is already struggling to reach the ideal FPS value... this is where it becomes significant, and VERY significant.

PS: Still you right, by itself it would not be reason enough to expend a bigger amount of cash in CPU, would not be the nowadays Intel CPU are giving better overall performance for almost the same cash compared to AMD. This is just fact, at least until Bulldozer proves itself it can recover a bit of the balance shift.

Addendum: Let's be clear, I'm biased against Intel but never from fanboyism. I don't let my preferences to cloud my judgement of facts. The only thing that keep me from Intel is it actively pursuing the TCG goals, not as blatantly as Sony but still hardly enough to give me the creeps. More than once it was caught trying to implement hardware DRM into the CPU and retreat afterwards. AMD too is signed to TCG, it needs to, but is not actively trying to implement anything "strange" to actual computing issues in their CPU, not yet at least.