Gigahertzes

B

Boa Constrictor

Limp Gawd
Joined
Nov 12, 2003
Messages
186
Ok, even the higher end AMD processors are only a little over 2GHZ speed wise. Top of the line Pentiums are over 3GHZ, however, in benchmarks, the GHZ # seems to make very little difference. Why?

Also, when you look at requirements for games, most of the time they mention GHZ #, so for a gamer, wouldnt more GHZ make more sense long term?
 
the reason AMD only go up to the low 2 gig range is because their chips are faster clock per clock than an intel chip. so more ghz is not always the best answer

i'm not sure exactly how AMD and Intel compare but a example might be you have an AMD rated at 1.8 and a P4 rated at 2.4 and essentially they are they same speed. AGAIN that is just an example those figures are not true to life.
 
Originally posted by Boa Constrictor
Ok, even the higher end AMD processors are only a little over 2GHZ speed wise. Top of the line Pentiums are over 3GHZ, however, in benchmarks, the GHZ # seems to make very little difference. Why?

Also, when you look at requirements for games, most of the time they mention GHZ #, so for a gamer, wouldnt more GHZ make more sense long term?
Click to expand...

Because Intel and AMD chips are different chip architectures. MHz comparisons only make sense within the same architecture, not between architectures.
 
That is one of the main reasons AMD changed the naming of their processors to make up for the differences of the architectures. The number designation of AMD relates to the speed ratings of the Intel chips for a more fair comparison to consumers.
 
The AMD chips have a higher IPC. This means they get more work done per cycle. As a result, they do not have to operate at frequencies as high as Intel chips in order to get the same amount of work done.
 
Different chip designs give different performance characteristics at different frequencies. These designs also limit the frequency that can be attained with a certain manufacturing process. If AMD could get their Athlon 64/FX to run at 3GHz, believe me ... they would!
 
While you guys are on a row iwth the gighertzes stuff :) I know this should probably be in mobile. How do the say Centrino(i know centrino is the makeup of the processor, mobo, and wirless, so whatever proc the centrino uses then) processors compare to a regular Pentium? Someone was saying a Centrino 1.6 relates to a Pentium 2.8 or soemthing? Can anyone clear this up for me?
 
Yeah the centrino CPU's (Pentium M) seem to be able to keep up with a P4 with a much higher clock speed. What speed desktop chip it is equal to is very dependant on what you are running especially as the mobile machines have a much slower FSB
 
ya i never did understand why amd seems lower than intel... what is my 2500 equivalent too?

-warsaw
 
i always thought that since the AMD's have 6 less stages to the CPU, they refresh the pipeline faster and get more per done per clock cycle. I'm pretty the the Pentium's 20 stage pipeline is there for ramping up clock speeds. The Pentium M's i believe have either two more or less stages than the Athlons, so they're low clock speeds can still compete (like the Athlons) with the Pentium's. I dunno, I'm prolly talkin out my ass, it is new years after all.
 
heres an idea, make an dual proc board for one intel and one amd proc. that would very interesting, but probably impossible, or at the bery least, very expesinve. just like the dual gpu radeon 9800 pro from sapphire.
 
Originally posted by fallsincluded
heres an idea, make an dual proc board for one intel and one amd proc. that would very interesting, but probably impossible, or at the bery least, very expesinve. just like the dual gpu radeon 9800 pro from sapphire.
Click to expand...

Yea that would be impossible because of the different architecture used on Intel and AMD processors and motherboards.
 
Originally posted by Shyne151
While you guys are on a row iwth the gighertzes stuff :) I know this should probably be in mobile. How do the say Centrino(i know centrino is the makeup of the processor, mobo, and wirless, so whatever proc the centrino uses then) processors compare to a regular Pentium? Someone was saying a Centrino 1.6 relates to a Pentium 2.8 or soemthing? Can anyone clear this up for me?
Click to expand...

Because the Pentium Ms (aka Centrinos) have larger on-die caches, providing higher performance at lower clock speeds.
 
Originally posted by skritch
Because the Pentium Ms (aka Centrinos) have larger on-die caches, providing higher performance at lower clock speeds.
Click to expand...

That does not even come close to fully explaining the performance delta, sorry.
 
The answer is this;

McClaren F1 or Porsche 911 GT, oh yeah everytime you want to go somewhere in the McClaren you have to go via Rio de Janiero.

Intel is the McClaren.
 
Originally posted by rolo
That does not even come close to fully explaining the performance delta, sorry.
Click to expand...

Ah, sorry. Then why didn't you enlighten us, being posessed of this knowledge I neglected to post?
 
Think about it this way. A father and his child walk together. They travel at roughly the same speed, but the child must move her legs more rapidly than the father to keep up. The Pentium 4 is the child because it runs at a higher frequency (takes steps more frequently), while the Athlon (XP or 64) is the father because he doesn't need to take as many steps per second to travel at the same speed (application performance).
 
some of these comparisons are terrible, i had no idea what the hell the car thing was about
 
Originally posted by MajorDomo
That is one of the main reasons AMD changed the naming of their processors to make up for the differences of the architectures. The number designation of AMD relates to the speed ratings of the Intel chips for a more fair comparison to consumers.
Click to expand...

I'm gonna have to sort of disagree with you there. According to AMD it's naming scheme is based on it's T-bird line. And is "supposed" to have nothing to do with Intel's procs.
 
Originally posted by Asian Fury
some of these comparisons are terrible, i had no idea what the hell the car thing was about
Click to expand...

That's OK, your daddy planted a seed in your mummys pee-pee and it grew to become you.
 
Originally posted by xonik
Think about it this way. A father and his child walk together. They travel at roughly the same speed, but the child must move her legs more rapidly than the father to keep up. The Pentium 4 is the child because it runs at a higher frequency (takes steps more frequently), while the Athlon (XP or 64) is the father because he doesn't need to take as many steps per second to travel at the same speed (application performance).
Click to expand...

I like this one^^^.

In addition, the reason the Pentium-M is comparable to Pentium 4s running at much higher frequencies is because the M is based on the PIII. The PIII had a higher IPC than the P4 does; this is why when the P4 1.3s and 1.4s came out, everyone was angry because the Tualatin (last gen PIII core) at 1.26GHz spanked the newer chips.

The Pentium-M performs so well because it has an even higher IPC than the PIII and the AthlonXP, runs at reasonably high frequencies, has an acceptable 400MHz FSB, and has a large cache. 'tis a marvel of engineering.

Anandtech.com (IIRC) had a review that showed the Pentium-M 1.6GHz comparing roughly with a P4-2.66GHz in business benches, IIRC, but got owned in the multimedia tests. AMD tends to perform well in business apps also.

edited to add: I didn't get the Rio de Janeiro analogy either, would you mind elaborating?
 
Originally posted by Mad_Pyro
<snip>

edited to add: I didn't get the Rio de Janeiro analogy either, would you mind elaborating?
Click to expand...

C'mon now mang The McClaren is obviously faster it just goes a circuitous route to get to the same place. :rolleyes:
 
Originally posted by skritch
Ah, sorry. Then why didn't you enlighten us, being posessed of this knowledge I neglected to post?
Click to expand...

Hah, sorry I was kind pressed for time at the moment.

Anyway what the other guy said was true. The P-M is based off the P3 and has a higher IPC in general. The P4, however, will still perform better for certain things (MPEG encoding, anyone?), and can be ramped to a higher clock speed. The P-M's biggest benefit is probably battery life while still maintaining really good performance. On the desktop you'd still want a really high clocked P4. Same price with better performance (albeit much higher power consumption).
 
Originally posted by rolo
Hah, sorry I was kind pressed for time at the moment.

Anyway what the other guy said was true. The P-M is based off the P3 and has a higher IPC in general. The P4, however, will still perform better for certain things (MPEG encoding, anyone?), and can be ramped to a higher clock speed. The P-M's biggest benefit is probably battery life while still maintaining really good performance. On the desktop you'd still want a really high clocked P4. Same price with better performance (albeit much higher power consumption).
Click to expand...

...and how does this invalidate my claim that the performance is due to the much larger on-die cache on the P-M? Larger cache = greater performance at lower clock speeds = lower heat generation and also lower power consumption = better battery life.
 
Originally posted by skritch
...and how does this invalidate my claim that the performance is due to the much larger on-die cache on the P-M? Larger cache = greater performance at lower clock speeds = lower heat generation and also lower power consumption = better battery life.
Click to expand...

The L2 cache gives some points to performance, but does not make the full gap between a 1.6 P-M and a 2.6 P4 (or whatever the comparison is).
 
Originally posted by rolo
The L2 cache gives some points to performance, but does not make the full gap between a 1.6 P-M and a 2.6 P4 (or whatever the comparison is).
Click to expand...

And, again, I'll ask for support for your claim.
 
Originally posted by skritch
And, again, I'll ask for support for your claim.
Click to expand...

likewise- it is NOT megahertz that matters so much in the performance difference, IT IS THE EFFICIENCY OF THE ARCHITECTURE!

:eek:
 
Originally posted by DeepFreeze
likewise- it is NOT megahertz that matters so much in the performance difference, IT IS THE EFFICIENCY OF THE ARCHITECTURE!

:eek:
Click to expand...

I didn't claim megahertz matters. I claimed the P-Ms perform like faster Ps due to the large L2 cache. The other person claimed the L2 cache doesn't account for the difference.

Since he is disagreeing with me, he's got the burden of proof. Have at it.
 
Look no further than the Pentium 4 Extreme Edition. It boasts a whole 2 MB more of internal, on-die cache, but do you see it performing several times faster than the 3.2C? By your logic it would...
 
Originally posted by xonik
Look no further than the Pentium 4 Extreme Edition. It boasts a whole 2 MB more of internal, on-die cache, but do you see it performing several times faster than the 3.2C? By your logic it would...
Click to expand...

No, I don't. Know why? It's not on-die cache. It's on-chip cache, i.e., L3 cache. NOT L2 cache. There's a SEVERE difference between the two.
 
Originally posted by skritch
No, I don't. Know why? It's not on-die cache. It's on-chip cache, i.e., L3 cache. NOT L2 cache. There's a SEVERE difference between the two.
Click to expand...

Skirtch. i said likewise, meaning i agree with you.
 
Skritch, there would be a significant difference if that was the case, but...

http://www.intel.com/design/Pentium4/prodbref/index.htm?iid=ipp_dlc_procp4phtxe+prod_brief&

Integrated 2MB Level 3 Cache
The 2MB L3 Cache is available with only the Pentium 4 processor Extreme Edition at 3.20 GHz. The additional third level of cache is located on the processor die and is designed specifically to meet the compute needs of high-end gamers and other power users. The Integrated Level 3 Cache is available in 2MB and is coupled with the 800 MHz system bus to provide a high bandwidth path to memory. The efficient design of the integrated Level 3 cache provides a faster path to large data sets stored in cache on the processor. This results in reduced average memory latency and increased throughput for larger workloads.
Click to expand...
(emphasis mine)
 
Uh oh, the burden of proof is now on you to explain why L3 cache is slower than L2 cache...

Seriously, why do you think that cache is what makes or break a CPU's performance? Certainly if Intel and AMD thought that cache was that important to CPU performance, they would invest heavily in embedded memory applications and throw out their sequential logic experience. Perhaps you should dig deeper into why a CPU performs as it does. Ace's Hardware, Ars Technica, and Lost Circuits are good resources for this kind of invesitigation.
 
Originally posted by xonik
Uh oh, the burden of proof is now on you to explain why L3 cache is slower than L2 cache...

Seriously, why do you think that cache is what makes or break a CPU's performance? Certainly if Intel and AMD thought that cache was that important to CPU performance, they would invest heavily in embedded memory applications and throw out their sequential logic experience. Perhaps you should dig deeper into why a CPU performs as it does. Ace's Hardware, Ars Technica, and Lost Circuits are good resources for this kind of invesitigation.
Click to expand...

Cache makes or breaks a CPUs performance because it reduces the CPus dependence on its largest bottleneck -- the throughput between the CPU and the system RAM.

L3 cache allows a slight speed improvement in accessing system RAM.

Since the L1 cache is faster than the L2 cache and the L2 cache is a lot faster than motherboard memory, increasing cache size and possibly adding a L3 cache between motherboard RAM and the L2 cache will increase performance. This is harder to quantify than bus speeds and multipliers, but increasing the L1 cache would reduce the number of times the CPU had to look to the L2 cache. Allowing a large L2 cache will likewise reduce the number of times the CPU has to look beyond the cache. Adding a L3 cache at motherboard speed may reduce the number of times main memory (had to be accessed, although there are other performance tradeoffs at this level.


In short, the L3 cache assists the CPU in accessing memory. It's hit whenever the L1 and L2 caches are filled. The L3 cache is MUCH slower than the L1 and L2 caches, and tends to run at motherboard bus speed, not CPU clock speed. So, any time you have to go beyond L2 cache, you have a CPU sitting there twiddling its thumbs, waiting for a return from memory access (which is deathly slow from the CPU's point of view).

Larger L2 cache means the CPU is fully utilized more often.

This is why (HINT HINT) the L2 cache in the high-end Xeons is large. It's also why they're so expensive (that, and adding large L2 cache decreases viable yield for a given CPU batch).
 
L3 cache is slower because of it's implementation. Each cache on your cpu is not made equal. The general sram model uses 6 transistors for each bit. This is quite large but it is also quite fast. The problem is the faster something has to operate in general the larger the area has to be. Cost of silicon is proportional to the area to the power of 6 (or something like that can't remember what my Digital prof said. But it is something huge). That is why L1 cache is so small in size. As you move further and further away from the cpu (L1->L2->L3) density increases so something must decrease and that is speed. To look at other cache in the cpu look at the TLB's. They are built on extremely fast cache faster then your intruction or data caches.
The caches all operatate on the same clock when on die (ie 3ghz) but that does not mean all data/instructions can be fetched from cache in 1 clock cycle.

Cache makes or breaks a cpu's preformance because cache even though L3 is slower then L1 it is 100x faster then access to DRAM or even worse 1000x faster then your mechanical/optical storage. Everytime there is a cache miss you must go to main memory. If main memory does not have what you want you get a page fault and must go to your mechanical/optical storage to read thus increasing even more your miss penalty for cache. P4 and AMD both use prefech algorithms to anticipate the data/instruction that will be needed next in the sequence. So for a large amount of the time the cpu is only working in cache which speeds it up. (this is in the magnitude of nanoseconds) While every miss will cause a stall for the cpu and that is in the order much greater sometimes in the milliseconds if mechanical/optical drives are required.
 
Cache being the "largest bottleneck" is a debatable statement, considering that the P4 EE has varying and usually marginal performance gains over a similarly clocked P4 3.2C. Looking at Barton vs. Thoroughbred comparisons, you'll see that the effect of a DOUBLED L2 cache lends a scant 2-3% performance gain at identical clocks.

And just how is the L3 cache much slower than L2 cache when it's on-die? There is very, very little difference in latency between the two cache levels when they are on-die, to the point that they could be effectively combined. Just look at a picture of the die layout and you will see a very small difference in trace length from the cache to the ALU or FPU, leading to very small differences in latency and resultant performance. Your reasoning would have held water when L3 caches were on-chip or even discrete, but now it all falls apart.

Finally, all current model Xeons have 512 kB of L2 cache, with varying levels of that forbidden L3 cache.
 
Originally posted by xonik
Cache being the "largest bottleneck" is a debatable statement
Click to expand...

Yes, it is. Which is why I didn't say it. I said that the largest bottleneck is data transfer between the CPU and system RAM. Reread that sentence.
 
But if "L3 cache allows a slight speed improvement in accessing system RAM," then wouldn't it too be a part of this awful bottleneck?
 
You must log in or register to reply here.
Back
Top