Part of the reason why Intel sucks... Poor game code.

[Diasper]

Interesting article at Extremtech discussing what lack of code optimizations negatively affect P4 performance.

Of course, if code was properly optimized AMD64 as well as P4 would benefit (albeit probably P4 would see a much larger % increase).

I guess this also alludes to several other debates. Namely,

How consoles are more optmized (or rather PCs under-optimized)
How game publishers are lazy - especially on the PC (patches people?)
How EA sucks?
Will MS's compilers be much better in Vista?


Also worth a read is their readers' comments

 

[Furen]

Developers normally optimize by hand rather than rely on auto-optimizers because they can sometimes break your code (and people complain enough about their buggy code without introducing new bugs). The K7 matched the current K8 in FPU performance but it sucked against the Northwood in games (and everything else... of course it was bandwidth starved and all but even a single-channel Sempron will spank any K7 in games).

 

[Markbnj]

I've never used strong compiler optimization on the critical parts, and never will.

 

[sonoran]

Doesn't surprise me really. Fact is the P4 has always had kicking floating point performance - *if* you use SSE/SSE2 to do the calculations. I don't think anyone's foolish enough to dispute that it comes in 2nd place when the code is heavy in base FPU instructions. Benchies tend to stress the latest/greatest (read SSE/2), whereas old real world code doesn't use it (and as that article shows even new code might not). That's why there tends to be such a disparity between benchmarks and apps. It'd be interesting to see benchmarks on a P4 with BF2 compiled with and without SSE2.

* Not speaking for Intel Corp *

 

[imported_Technomancer]

Ok, part of the problem with that is the fact that there isn't a single Pentium 4. There was Willamette (sp?), Northwood, and I beleive that the last ones used Prescott architectures. Each architecture is very different; optimizations for a Northwood will reduce a Prescott's performance, and vice versa.

The main issue is that all console developers get to develop on the same hardware (generally) no matter what. Thus optimization is "easier". But, for PC's, just try and write an optimized program variant for each CPU/memory/graphics card/HDD available on market. They're fortunate just to get it to work at all.

BTW EA does suck, but that's not the point. They just release the same game each year with only cosmetic changes to the title.

Vista- I have my own problems wityh it, I dont want to get into any online flame wars because they waste my time. It's also why I don't post much anymore.

 

[Zebo]

Right, one game he reversed compiled (illegally i might add) does'nt use specific optimizations so all games don't.:roll:

Nevermind the whole premise is moot since A64's fully support Intel's instructions so whatever gain intel got from them so would AMD.. Therefore P4 would still suck as a relative measure to thier competition even with implementations.

Face it p4 just sucks, or should I say long in the tooth..dated. Pentium-M does'nt have this gaming problem.

BTW - Why is'nt anyone crying about AMD's optimizations not being in there like 3dnow(+)?

 

[Fox5]

Originally posted by: Zebo
Right, one game he reversed compiled (illegally i might add) does'nt use specific optimizations so all games don't.:roll:

Nevermind the whole premise is moot since A64's fully support Intel's instructions so whatever gain intel got from them so would AMD.. Therefore P4 would still suck as a relative measure to thier competition even with implementations.

Face it p4 just sucks, or should I say long in the tooth..dated. Pentium-M does'nt have this gaming problem.

BTW - Why is'nt anyone crying about AMD's optimizations not being in there like 3dnow(+)?

Not true, P4's can see up to 50% performance increases from SSE, Athlons usually don't see more than a 10% increase, and sometimes even see a slight decrease.
Not sure how 3dnow+ affected athlon's performance, but it did give like a 50% boost to their k6 cpus. Part of the design of the Athlon focused on fixing this....to make a cpu that performs good without SIMD or custom coding since AMD wasn't a big enough part of the market to demand it. (games can be optimized for athlons though, just take a look at the unreal series)

 

[Griswold]

Originally posted by: Fox5

Originally posted by: Zebo
Right, one game he reversed compiled (illegally i might add) does'nt use specific optimizations so all games don't.:roll:

Nevermind the whole premise is moot since A64's fully support Intel's instructions so whatever gain intel got from them so would AMD.. Therefore P4 would still suck as a relative measure to thier competition even with implementations.

Face it p4 just sucks, or should I say long in the tooth..dated. Pentium-M does'nt have this gaming problem.

BTW - Why is'nt anyone crying about AMD's optimizations not being in there like 3dnow(+)?

Not true, P4's can see up to 50% performance increases from SSE, Athlons usually don't see more than a 10% increase, and sometimes even see a slight decrease.

Where did you get these numbers from?

 

[BrownTown]

Its pretty well known that the pentium 4 works best with well optimized code. Its clock speed is much higher, so if it can keep its pipes full it can beat the A64. However, in branchy and and depentant code the P4 really suffers. It takes a P4 longer to execute an instruction then the A64, so if you are always stalling and waiting for data the P4 will suffer alot more.

Lots of people talk like netburst is a crappy design, but it really isn't, Intels problem was their process technology putting out too much heat to allow them to clock the P4 as high as it should be able to go. Personally I'd rather see Intel going back to a 20 stage pipeline like the Northwood instead of all the way back to 14 which really limits your maximum thoroughput.