AMD And Intel?

[s3nate]

Thanks everyone for their advice, really helped. I think that im just gonna get an Intel P4 3.4ghz =)

Thanks,
S3nate

[yourdeadin]

[i] Anyways, for a cheaper and still very powerful rig, consider AthlonMP systems. If you want to try that, I'll put up a list of what to get to make the process as painless, cheap, and powerful as possible. Although overclocking is required, and if you mess up your investment could go up in smoke... Not for the faint of heart, or someone who messes stuff up [/B]

yes especialy if you are in for a coparate world , you would save a lot of money with having to cut down on the processing speed . i'd say AMD is far better than intel (personal opinion) intel advertises but AMD doesnm't, much like "windows or lnux" as far as AMD 64 is considered we do not have enough applications built for 64 bit computing, most of them are a a briged gap between 32 and 64 bit application

[I]Oh the hell with 64 bit. I'll take a dual Intel Xeon 3.06 GHz box with 4 GBs of ECC RAM over some AMD 64 crap any day.

there are actually no 64 bit application built for it ( exceptions are there !) but no dedicated 64 bit os ( not sure yet)

www.pricewatch.com would be the place to start.

It doesn't matter how fast your processor is if the RAM and MoBo can't keep up with it, so if you want speed, get a 10,000 RPM SCSI HD, a GOOD MotherBoard, FAST RAM, a very good Video card, something like an Nvidia GE Force FX (ATI blows) and don't get a crap fan either. Make sure everything is cooled properly

could be more right , you need enough ram to support you processer speed
is it just me or ,as i can't find the intel xenon price

AMD Pros:
- Generally much cheaper than equivalently performing Intel processors
- Lower memory latency, especially in the Athlon 64s, which feature an on-die memory controller
- Generally better gaming performance
- First to expand to AMD64 (x86-64), which will allow you to address >4GB of RAM without crappy workarounds, and are MUCH MUCH better in SMP (Multiple Processors) than Intel's Xeons.
- AMD64 processors will also perform 64-bit mathematical operations in half the time a 32-bit processor would. This benefit is felt most in floating-point math.

I'm going to avoid going into detail about how wrong certain posters are, but suffice it to say, if you want Price:Performance, go AMD, if you want raw performance, not caring about price, go Intel. This may not remain true forever, but it is true now, and will be true for some time.

could be more right !

[gore]

No 64 bit OS???

x86 has Linux and a few other OSs which are 64 bit, and other platforms like Sun have had 64 bit OSs for years.

[cacosapo]

just adding some comments

Pentium series, celeron, 386, 486, all AMD XP and Duron (and via cpus too) are the same archictecure - IA-32, patented by Intel. The basic diference between them is the APPROACH to implement the archictetures. Its like cars - everyone has the same base technology, but they can have very diferent performance.

AMD moved out from Intel archicteture and developts its own archicteture - x86-64 - now named AMD64

Intel, due the end of IA-32 (no more scalability) created a new one - IA-64 - there where's itanium belongs

Higher core clock means faster math processing (Intel)

not true everytime. Some risc processors have clock from 1ghz to 2ghz and FASTER FASTER than any Intel 3 ghz

Generally higher memory bandwidth, depending on the memory used.

true, but its not a real advantage. Intel needs that because its processor really need that. AMD went to another way to avoid increase speed of FSB and memory

MUCH MUCH better in SMP (Multiple Processors) than Intel's Xeons.

dont think so, and SMP xeon processor are far away more popular on companies than AMD SMP

First to expand to AMD64 (x86-64), which will allow you to address >4GB of RAM

even on IA-32 you can have more than 4GB of RAM. IA-32 provide ways to an 32 bit O.S. address more than 4GB of real memory. 64 bits hardware allow that private address space ("virtual memory" to expand more than 4GB.

[chsh]

Originally posted here by cacosapo
not true everytime. Some risc processors have clock from 1ghz to 2ghz and FASTER FASTER than any Intel 3 ghz

Firstly, I wasn't discussing RISC processors, which I thought was blatantly obvious, and secondly, yes, it is still true. The reason RISC processors at 1GHz are very fast is because they do much more work per clock. Most RISC processors have a 4 or 6 stage pipeline, while most modern CISC processors (Intel-compatible desktop processors) have 12-30 stage pipelines (depending on the manufacturer). If you look at the raw GHz speeds of the processors, you will see an Athlon-64 (a 12 stage pipeline) running in 32bit mode at around 2.2GHz performs mostly on par with Intel 3.2GHz processors. However, for mathematical calculations that fit inside L2 cache, the Intels will generally win because there are shorter paths that simple math and such can take. The pipeline length on Intel processors is variable, and in some instances it can be as low as 15 I believe (for simple math), which means that while an Intel processor will have 3 more stages (3 more Hz needed for one instruction), it is still able to be clocked so much higher that it is still much faster.

true, but its not a real advantage. Intel needs that because its processor really need that. AMD went to another way to avoid increase speed of FSB and memory

Yes, it is a real advantage, and you have that backwards. AMD has been steadily upping their FSB/memory speeds, Intel's had the same for some time now. Intel processors benefit from more cache to decrease memory latency, while AMD processors have exhibited more benefit from higher speed RAM.

dont think so, and SMP xeon processor are far away more popular on companies than AMD SMP

I wasn't discussing popularity, I was discussing useability and wellness of design. The popularity of Intel's server processors is rapidly declining now that AMD has had a few years to build its reputation in the server industry. Anyway, my point about the AMD's design being better was this: Intel's processors work off a shared bus. So for example, if you have 4 Xeons in a box, they all share a 1000MHz path to RAM. With Athlon MPs and Athlon-64s, each processor is given an independant path to RAM, meaning they get the full benefit of their FSB. An example: In a Quad Athlon 64 system with 400MHz DDR RAM, each processor is given 400MHz access speed to RAM. In a Quad Xeon box with 1000MHz RDRAM, each processor is given 1000/4 or 250MHz. The athlons also benefit from lower overall latency.
Relevant Linkage:
http://www.anandtech.com/IT/showdoc.aspx?i=2205 << Very recent - Opteron 250 vs Xeon 3.6
http://www.anandtech.com/IT/showdoc.aspx?i=1982 << Somewhat Recent - 2x & 4x Opteron 848 vs 2x & 4x Xeon 3.2s
http://www.anandtech.com/IT/showdoc.aspx?i=2173 << Somewhat Recent - What Anandtech uses.
http://www.anandtech.com/cpuchipsets...oc.aspx?i=1483 << Very Old - Athlon MP 1.2 vs Xeon 1.7s.

even on IA-32 you can have more than 4GB of RAM. IA-32 provide ways to an 32 bit O.S. address more than 4GB of real memory. 64 bits hardware allow that private address space ("virtual memory" to expand more than 4GB.

This is incorrect. IA-32 did NOT support more than 4GB of RAM, so Intel added what they call PAE (Physical Address Extension) to the Pentium III and higher processors to support (I believe) up to 39 bits of addressing space. The Athlon-64 is not a true 64 bit processor in its addressability, it is capable of addressing only 48 bits (darn, it means 281 TeraBytes instead of 1 844 PetaBytes). A 64-bit processor also does more than just allow you to address up to 64 bits, it does math larger than 32 bits in half the time, and in the case of the x86-64, there are double the registers of x86.

Here's some relevant reading:
http://www.intel.com/support/process...1617-prd46.htm
http://support.microsoft.com/default...268363&sd=tech

[cacosapo]

I wasn't discussing RISC processors, which I thought was blatantly obvious, and secondly, yes, it is still true.

sorry, csch, you missed my point. i gave RISC processor as an example about that not allways higher clock speed means higher math process. there is a plenty of other hardware schemas that allows speed up on operations, such as math.
I was trying to generaly speaking about comparing architectures to thread starter. IMHO, to compare Intel and AMD i think some basic knowledge is good too.
[/QUOTE]

Yes, it is a real advantage, and you have that backwards. AMD has been steadily upping their FSB/memory speeds, Intel's had the same for some time now. Intel processors benefit from more cache to decrease memory latency, while AMD processors have exhibited more benefit from higher speed RAM.

Again, you stickied on IA-32 architecture, while i was thinking in general. But If you look at battle between Pentium and Athlon, you can see that Intel almost lost BECAUSE its processor ar hunger for high bandwidth, and AMD could beat Intel most of the times using a lower FSB clock.

I wasn't discussing popularity, I was discussing useability and wellness of design. The popularity of Intel's server processors is rapidly declining now that AMD has had a few years to build its reputation in the server industry. Anyway, my point about the AMD's design being better was this: Intel's processors work off a shared bus. So for example, if you have 4 Xeons in a box, they all share a 1000MHz path to RAM. With Athlon MPs and Athlon-64s, each processor is given an independant path to RAM, meaning they get the full benefit of their FSB. An example: In a Quad Athlon 64 system with 400MHz DDR RAM, each processor is given 400MHz access speed to RAM. In a Quad Xeon box with 1000MHz RDRAM, each processor is given 1000/4 or 250MHz. The athlons also benefit from lower overall latency.

sorry, i didnt explain my point of view correctly. When i talked about "popularity" i was meaning that market thinks that Intel SMP archicteture is better.
But lets discuss about shared bus and dedicated bus.
Shared bus are pretty common on SMP because:

1) easy to do
2) we can maintain cache coherence better - its a nightmare to sync cpu caches on a SMP structure - after 16 processors, sometimes we start to LOOSE performance isntead gain thruput. So most archicture has shared buses for all processors or at least a group of them.

This is incorrect. IA-32 did NOT support more than 4GB of RAM, so Intel added what they call PAE (Physical Address Extension) to the Pentium III and higher processors to support (I believe) up to 39 bits of addressing space.

so its what ive said. Intel add extensions (PAE) to allow IA-32 processors to addres more Real Memory, You can run a IA-32 machines with 8 GB(or more) of RAM and use it. A far i know (but i maybe be wrong) IA-32 aplications, even with PAM, cann create an address space greater than 4GB.
Those extensions were created on dinossaurs (IBM Mainframes) a long time ago - most of them are stucked on 31-bit addressing.

[chsh]

Originally posted here by cacosapo
Again, you stickied on IA-32 architecture, while i was thinking in general. But If you look at battle between Pentium and Athlon, you can see that Intel almost lost BECAUSE its processor ar hunger for high bandwidth, and AMD could beat Intel most of the times using a lower FSB clock.

IA32 has nothing to do with the general architecture you were discussing AMD and Intel directly. In the future, don't directly refer to them if you are "speaking in general", because it doesn't make it speaking in general if you do, does it?

sorry, i didnt explain my point of view correctly. When i talked about "popularity" i was meaning that market thinks that Intel SMP archicteture is better.

No, they don't, which is why AMD has gained so much popularity in the last few years. Intel held the lead because they were really the only "x86 server chip" game in town.

But lets discuss about shared bus and dedicated bus.
Shared bus are pretty common on SMP because:
1) easy to do
2) we can maintain cache coherence better - its a nightmare to sync cpu caches on a SMP structure - after 16 processors, sometimes we start to LOOSE performance isntead gain thruput. So most archicture has shared buses for all processors or at least a group of them.

And yet, if you read the articles I linked AMD is maintaining BETTER cache coherence than Intel, using a non-shared bus architecture.

so its what ive said. Intel add extensions (PAE) to allow IA-32 processors to addres more Real Memory, You can run a IA-32 machines with 8 GB(or more) of RAM and use it.

No, they did not., they added it to SPECIFIC processor lines going forward. Your statement of "You can run a IA-32 machines with 8GB(or more) of RAM and use it" is incorrect because The 386, 486, 586/Pentium, and 686/PentiumPro/Pentium II are all IA-32 processors, and yet all lack PAE. Additionally, applications are stuck addressing 4GB of RAM, however you are able to run multiple applications that can address more. I encourage you to read the technical documentation I linked in my last post before responding, as it will confirm and further explain what I have said here.

[Tim_axe]

Shared busses + ( Bandwidth Hungry Processors * Multiple Processors ) = Electrical Engineering Problems.

Until very recently, with Intel's new Noconoa cores, Xeon's have been stuck at a 533MHz FSB. That isn't to say that the adventuring overclocker can't push it past that. It is just saying that Intel realized they had a problem of operating at an 800MHz bus (really 200MHz Quad-Pumped, but still fast) with two processors because at those speeds an error is fatal. (Which is why until recently, 800MHz was only for their single processor P4's) It is also much harder to get two chips running on an 800MHz bus because of some electrical issues that I don't throughly understand but know exist. A proof that this issue does exist, Intel supports their 800MHz Xeon processors only for dual configurations. Before that, 533MHz was as fast as their Dual Xeon processors would go. Their processors for 4+ CPU's are all at 400MHz. With how bandwidht hungry we've come to see Intel's processors to be, we know they'd want more bandwidth for more processors. But their product lineup is the opposite of that -- something is preventing them from doing that. And it is these electrical/engineering issues. We know Intel will give the better wafers of silicone to become more expensive processors, and that they don't have any faster FSB DP systems shows that Intel has hit some sort of wall. Maybe in the future that will change a bit, judging from how the new cores operate at higher speeds...

http://developer.intel.com/products/...xeon/index.htm <- Intel's Xeon Processors & Specs


So developing for a single bus can infact be more difficult. Think of it like adding another hard-drive to your IDE channel. That is now becoming obsolete because of the performance issues associated with this, and is why SATA has become so popular. (PATA = Single Bus, SATA = Multiple Busses) The market is generally realizing this change, and AMD's solution is now competing with Intel's big name to become more widely used.

[VictorKaum]

Originally posted here by gore
Oh the hell with 64 bit. I'll take a dual Intel Xeon 3.06 GHz box with 4 GBs of ECC RAM over some AMD 64 crap any day.

It doesn't matter how fast your processor is if the RAM and MoBo can't keep up with it, so if you want speed, get a 10,000 RPM SCSI HD, a GOOD MotherBoard, FAST RAM, a very good Video card, something like an Nvidia GE Force FX (ATI blows) and don't get a crap fan either. Make sure everything is cooled properly.

Stay away from integrated hardware, it's crap. Get a Sound Blaster sound card, a good NIC if you use Cable or DSL, and no one at this place is going to make me believe a better Intel processor exists than the Xeon.

If you go for dual processors, actually use an OS that can utilize this. BeOS was great for this, but Windows.... Heh, not to good with it.

Gore come on...

"the AMD 64 crap" is not aiming at the Xeon market, the Opteron is. I'm still convinced of the power of the Xeon beasts so I'm not going to say that the Opteron is better or faster however when budget comes in mind...
So compare the Athlon64 with the recent Pentium4 releases.

About speed: even better is a 15krpm SCSI hdd in a raid 0

hmm 'ATI blows' ??? The X800 from ATI is at the moment the worlds fastest VPU
Ever compared a RADEON X800 XT PE 256MB DDR with the new Geforce ones? I'm not so sure ATI is that bad...

[gore]

Install something other than Windows. (Linux, Free BSD) and play Unreal Tournament.

From what I've seen it's a nightmare making the drivers work on ATI. Nvidia however supports Free BSD and Linux with a driver they make and it allows complete 3d support.

I don't care what the Xeon has for competition, I said I'd rather have one than something else. ATI can be as fast as it wants, but when I do an update on SUSE and it grabs Nvidia drivers for me, and they actually work and build a Kernel for me, then I'm not switching to ATI so I can play UT 12 frames faster. If it doesn't lag, I'm happy, I don't really care how many frames it is.