best processor?

[Web Gecko]

bhome83 ,

Stability I think is key here. It makes sense to pair up an Intel processor with
an Intel chip set. If you are dealing with a P4 and RDRAM this is important
(have DDR chip sets for the P4 arrived yet ?) but then it's also important with fast AMD processors as well.

Stability as far as all your m/b and other components is important as well.

What interests me is how many core bugs/defects are actually slipping through the quality control process at Intel and AMD. The new processors are so bloody complex that to work at all the quality control process must be fine tuned and intensive. With multi-millions of transistors and shrinking micron/dye processes it's a miracle that these things even work.

In the old 8/16 bit micro days people often talked about undocumented instructions/features (bugs/defects ?) but I didn't see that many write ups about core defects. I wonder as regards the processor if this is the predominant source of crashes/instability with Windows and other applications software.

A networking consultant in Massachusettes (who I keep in touch with) tells me
he always goes for the tried and tested Intel platform for his business clients.
They require high levels of performance and reliability. He says his clients' Intel platforms go weeks without crashing/blue screening whereas he does not speak so highly of AMD in this regard (relegates AMD to the budget/home market).
I can understand his conservative approach especially where providing quality to his clients' is concerned.

Of course we all know that AMD are providing high class and quality opposition for Intels' Celeron, PIII and P4 processors (Duron, Tbird and XP). AMDs' product is very good.

But this is the main point I want to make. My consultant friend in MA says that he believes that most of our Windows and apps development and testing is done on
Intel platforms (interested to hear what others think) so we aren't necessarily screening/testing our software with enough AMD processor cores. It is of course in MSes and the other developers interests to ensure that major new software development and testing is done on as many Intel and AMD processor cores as possible.

I think that most of the crashes/blue screening and general instability we see on current systems may be due to this. Of course there are probably some core bugs/defects slipping through. These can't be major as quality/reliability and product image/market share would be affected too greatly.

I think this issue can also apply to other programmable chips on the m/b like your chip set etc..

Here is a question for someone. Are there any good sites out there with info/write ups on the current processor cores out there (Intel and AMD)
and pros and cons (plus bugs /defects) of these cores ?

Interested to hear what others think.

chou

The Web Gecko

[Toaster]

Hello there mister lizzard,

All CPU cores have some sort of "patchwork" to some degree.
While a site where these could be explored probably exists, I am not aware of such. I tend to seek out these and other "undocumented" goddies on my lonesome and with the help of others. Often (Intel does this), instructions are added/deleted during any given production run for a myriad of reasons. Often, cost of production is a chief reason. Other times, its sheer R&D to trial run certain ideas and concepts. As far as the chipsets, I only work with Intel specific sets to any degree. I don't/won't do VIA because what I did see didn't please me.
Intel on the otherhand has numerous undocumented registers in the BX and 810 chipsets. The BX is where I focused a great deal of my time.
I found in the BX chipset, the beginnings of RAMBUS memory implementation.
Seemed Intel knew that memory access speeds were staying static while faster and faster CPUs were being made.
A discovery me and a friend made showed great promise. This was a memory interleaving function that was built in but not made available through BIOS/jumper settings. This was when the system board had EQUAL pairs of memory slots. This most often worked with system boards sporting 4 sockets. When each socket was populated with alike modules (4-64MB/128MB), a register within the chipset can be "tripped" to utilize memory bank interleaving.
This resulted in a gain of 22% of memory access rates beyond nominal "good" figures. When we forced the register and a "latency" register to "1", memory speeds roze by nearly 40% without an impact on reliability.
These numbers approach multi-banked RDRAM figures under "ideal" settings. This is VERY fast memory moves that VIA can't come close to.
There are registers in both CPU and chipsets. Some are QUITE good and some downright disaterous. Many are set aside for "developmental/diagnostics" reasons and thier use is not recommended and often not supported by the system board.
Only the ASUS P2-B/F support these features to my knowledge.
Rumor has it the ABIT is the same but not the ABIT I have. (BH-6)

[Carl Price]

Originally posted by LawyerRon Well folks all I can tell you is AMD must be doing something right because about 90% of ALL new systems being built by the members at this forum and others are AMD. Further, they all seem to be satisfied with them. I'm a big fan of Intel but I think they've conceded the "enthusiast" market to AMD.

Ron, you may be right, but I like to think that the only people who truly want AMD systems are the people like us who can deal with the problems. I build Intel boxes because I can't afford to see each system 3-4 times for problems after it is sold. When you have to go to the "mechanic" with every problem you don't want any, especially if you are the "mechanic" and you are not getting paid extra for the job.

[Electron]

Going on the same quote from LawyerRon in Carl's reply above, i believe Intel have lost out at the moment because of the P4 and expensive Rambus. Over here in England most of the pre-built boxes are AMD and the general population are buying pre-builds being unable or unwilling to build their own. Also prices are more expensive over here and there is no way of competing with the big manufacturers.
Unless you understand about the different Mobo's and CPU's, how they perform, different types and speeds of RAM, on-board this and that, upgradability etc. etc. people are just looking at processor speed and price, for which the AMD will win.
Hopefully, this will change with the new chipset boards for the P4, when they arrive, which will probably be a few months after they are available in America. Then there might be more competition.

[Web Gecko]

Hey Toaster,

What you said about your discoveries with processors/chip sets was interesting.

How do you ID these undocumented registers etc.. ? Do you feed in a binary sequence into the register/operand area of your opcode or what i.e. which is outside the normal range for documented registers ?

How do you know you are dealing with undocumented registers ?
The memory interleaving sounds cool.

Interested.

cheers

The Web Gecko

[Toaster]

Hello again mister lizzard,
I use the old "push" "pull" trick of yore.
In the PC, "everything" has/uses an address. Thats how the lil mothers work.
I "push" values and/or "pop" values. This is an old school method but one I'm quite comfortable with. CPU registers are a bit different and have to be handled with care other wise a reset is often initiated.
If you are really intersted, Intel posts on thier web site thier "white papers".
This lists many addresses and register functions. Whats interesting is the ommisions and gaps that exist. I like to explore the gaps by plugging values into these registers. This MUST be done in a DOS or Unix environment.
In the Winderz environment, the CPU is often working and its common to initiate a "stall" which is more often then not fatal to the application.
I do the majority of my exploring in "X86" mode then "enhanced" or the more extended functions. On the chip side, I and others modified an existing register mod proggie so that what we call as "bad" registers are not tripped.
The good thing about Intel X86 based CPUs is that thier function is similar.
Unless one enables "enhanced" or "386" modes, the chip is just a fast 8088.
Its the "protected" modes where the trickery shows up and often many pitfalls or "traps" we like to call them. I think andmany others that Intel devised numerous "do nothing traps" to thwart what we are doing.
It is VERY easy to push a register and hang the system to the point the plug has to be pulled. I have "popped" a few BIOS chips in my time as well going down the learning curve.







Why folks flock to AMD?
Cheap...period.
You folks buy with your wallet, plain and simple.
Like Carl says, folks buy them because they like the never ending headaches.

[Terrorbyte]

Great to see this debate still going on, for the thread starters info, Should we compare a 1.9 g AMD and an equal value 1.9 g Intel cpu you will find the following;
The AMD is 15% faster on overall system performance.
The AMD is 19% faster with general office application performance.
The AMD is 13% faster with digital Media performance.
The AMD is 14% faster with 3D Gaming and framerate performance.

So the choice is fairly simple, depends on the choice of m/b, HDD and Video card, as this can make a big differene to the whole packages performance no matter what cpu is in it.

Yes I am biased, I love my XP.