OPTION BASE question (na_th_an?)

[TheBigBasicQ]

Thats precisely the problem. None of the apps/OSs that I have seen are coded using Intel's new instruction set. If someone actually did that then their processors will be just as fast as AMDs.

Zack, HT is not a new technology. It has been used in servers for quite sometime now. It basically fools the OS into thinking that there are two processors instead of one. Hence it increases its efficience only if the OS is optimised for it.

Take a look at this:
http://www.intel.com/technology/hyperthr...d=sr+hyper&

Are you lazy? Just use google =P.

[KiZ]

Sorry to short post and all, but Aga, why on earth is this in the "Newbie help" section?! I wouldnt exactly call you a newbie!??!

[adosorken]

Depends on the type of application you're running as well. AMD's K7 blows the snot out of Intel's P4 for all legacy applications, and clock-for-clock, is somehow superior than the P4 in "modern" applications as well.

The reasons for the P4's failure on legacy code is what they cal the "8 Deadly Design Flaws", which are:

Small L1 data cache (only 8K!)
No L3 cache (cut from the design)
Decoder is crippled (they killed the multiple decoders and went back to a single decoder)
Trace cache throughput too low
Wrong distribution of execution units
Shifts and rotates are slow (they eliminated the barrel shifter and replaced it with a shift/rotate execution unit, which takes many cycles to work, as opposed to the barrel shifter which works on a single cycle)
Fixed the partial register stall with a worse solution (partial register access now relies on the new shift/rotate unit, making partial register accesses slower than L1 memory)
Instructions take more clock cycles to complete

For example, an indirect call through a general purpose register, common when making member function calls in C++, now takes about 40 clock cycles on the Pentium 4. Compare this to only 10 to 14 cycles on P6 family and AMD Athlon processors. Even at the faster clock speed, the Pentium 4 function calls are slower overall. Similarly, Windows API calls, which call indirectly through an import table, are now slower. Several Windows APIs that I tested literally took 2 to 3 times the number of clock cycles to execute on the Pentium 4. This is because not only do all the internal function calls within Windows take longer, but you have to remember that Windows 2000 and Windows Millennium are compiled using C++ compilers that optimize for Pentium III and Athlon processors. So as I mentioned at the beginning, until such time as most Windows code is recompiled using as-yet-non-existent Pentium 4 optimized C++ compilers, the performance of Windows applications will be terrible on the Pentium 4 processor.

[TheBigBasicQ]

Whoa! That means if the latest AMD processors would run at 3.2GHz they would kick P4's A$$. And whats the idea of having just 8K of L1 cache while having 512/1024K of L2.

[LooseCaboose]

You forgot to mention Hyperthreading. Its one of the best ways of improving CPU usage and efficiency.

There are many ways to improve cpu effecieny. I mentioned caching and pipelining because they are two of the most common and useful speed improvements. A good caching system on a cpu can mean that 99% of all memory accesses are done through the cache. That is a huge improvement in speed. Hyperthreading is only just starting to become mainstream in processors.

Ah...so basically Intel is lazy? Or does AMD have a copyright on those techniques?

I was talking about processors in general, AMD and Intel (as well as other cpu manufactures) use these techniques to improve the speed of their processors. Each of these improvements requires some number of transistors to implement them. The design process of a cpu involves working out which features to use, how big to make the caches, etc.

No L3 cache (cut from the design)

Thats not a fault of the cpu. L1 caches are internal to the cpu, L2 caches are in the cpu package and L3 caches are located on the motherboard separate from the cpu.

And whats the idea of having just 8K of L1 cache while having 512/1024K of L2.

Basically the faster memory is and the closer it is to the cpu the more expensive it is. L1 cache sizes are also limited by the number of available transistors in the cpu as I mentioned above. Typically machines have around 4-64k of L1 cache, 256-2048k of L2 cache and a few mb of L3 cache.

[TheBigBasicQ]

So, if what you say is true then why isnt Intel implementing larger cache?

[LooseCaboose]

So, if what you say is true then why isnt Intel implementing larger cache?

Possibly because they are focusing on other aspects such as the pipeline (some Intel processors have 15 stage pipelines). Intel are also trying to keep the clockspeed of processors up, this doesn't necessarily make for the fastest processors, but it is simple marketing strategy. Which processor do you think the general population will go for, a 4ghz cpu or a 3ghz cpu that has a huge L1 cache? This is why AMD stopped naming their processors after the clockspeed, because they are trying to get the general public to understand that clockspeed isn't everything.

[TheBigBasicQ]

Intel's marketing strategy works for non-technical users. But users like me or you who wont fall for it.

But you gotta check out intel's high end processors with their mobos. 800Mhz FSB coupled with HT is a huge plus point for Intel while AMD's max FSB is just 400Mhz and no HT =(.

[na_th_an]

Well, that HT thing only works with specific applications, such as server applications. Go to the AMD site and watch the benchmarks. With general use applications, HT Pentium IVs perform even worse than non-HT Pentium IVs. Think about it, not many applications use multiple threads, but just one. HT is for processes that launch many threads. You can have several processes running, but HT won't do anything there. Threads <> Processes, you know.

http://www.amd.com/us-en/Processors/Prod...50,00.html



---

Now this is something that tells much for me: Read this page: http://www.amd.com/us-en/Processors/Tech...59,00.html

AMD lists the cases where XPs are better than PIVs, but it also lists the cases where PIVs are better. Go to the Intel web page and you won't find any comparatives.

[KiZ]

Bah. I hate the way AMD uses those common marketing techinques to make Pentium look rubbish. They make the scale very large so it looks like the p4 runs a lot slower than the AMD proccessors, and they use dark, depressing colours for the pentiums and nice green hues for the AMDs.

Im not saying that AMD are crap or anything, though. I cant decide which processor im gonna get in my next PC...