![[H]ard|Forum](/styles/hardforum/xenforo/logo_dark.png){kind=logo}
Can someone explain MHz/GHz myth to me
- Thread starter Ke0
- Start date
DarthWombaT
[H]ard|Gawd
- Joined
- Jan 7, 2002
- Messages
- 1,663
As I understand it, some processors are more 'efficient' and get more done per cycle then other once. Hence the idea that the frequency is not an adequate depiction of the actual performance, and the idea behind Performance Ratings.
(cf)Eclipse
Freelance Overclocker
- Joined
- Feb 18, 2003
- Messages
- 30,027
the best example i heard of is regarding a pickup truck and a semi.
the pickup truck can go faster than the semi, but it moves less during each trip
the semi, while being slower, can move more each trip.
this way, intel is the pickup. it runs faster, but each trip gets less done. amd is the semi. it is slower, but because it can do more each trip, it can have better performance at slower speeds.
not a great explaination of it, but it's a start
the pickup truck can go faster than the semi, but it moves less during each trip
the semi, while being slower, can move more each trip.
this way, intel is the pickup. it runs faster, but each trip gets less done. amd is the semi. it is slower, but because it can do more each trip, it can have better performance at slower speeds.
not a great explaination of it, but it's a start
Party2go9820
2[H]4U
- Joined
- Aug 11, 2000
- Messages
- 3,695
Just think of it like this - you've got a big pile of dirt that needs to be moved via wheelbarrow. So you load it up and move the first load, dump it and start loading the second. The faster you go the faster the work gets done (more mhz/ghz's).
Now your buddy shows up with the front end loader and takes a big huge bite out of the pile. He's moving slower than you (less mhz) but he got more done in the time available.
The Ghz/Mhz myth is right there - just because you're moving fast doesn't mean you're doing the most work. Getting the work done is the most important thing. Hence why the new AMD chips might only be running at ~2000mhz but performing the same as an Intel chip running much faster.
<Doh! I'm just a slow typer>
Now your buddy shows up with the front end loader and takes a big huge bite out of the pile. He's moving slower than you (less mhz) but he got more done in the time available.
The Ghz/Mhz myth is right there - just because you're moving fast doesn't mean you're doing the most work. Getting the work done is the most important thing. Hence why the new AMD chips might only be running at ~2000mhz but performing the same as an Intel chip running much faster.
<Doh! I'm just a slow typer>
Which works out great for AMD... Intel has hit a roadblock getting to 4 ghz stable on stock... For whatever reason, I believe energy loss(aka heat) in their 90nm process has stopped them...
If Intel can change their architecture (which I believe their doing) to do more work per cycle and keep the mhz up at the same time, then can retake the crown... I've been out of the game for a while, so I don't know... but I believe this is why their changing to a numbering scheme similiar to AMD's.
Intels realizing speed doesn't always matter...
If Intel can change their architecture (which I believe their doing) to do more work per cycle and keep the mhz up at the same time, then can retake the crown... I've been out of the game for a while, so I don't know... but I believe this is why their changing to a numbering scheme similiar to AMD's.
Intels realizing speed doesn't always matter...
As a point of reference - P3s were were pretty much tied with Athlons at the same clock speed when the P4 came out. Older 1GHz chips were killing 1.4 and 1.5GHz p4s. This is just general architectural efficiency.
Another - PowerPC chips (like the G4 and G5 in newer Macs) have a kick-ass vector processor with Altivec. G5s running properly tuned software that fully utilizes Altivec can beat or meet a P4 at twice the speed (running good SSE/SSE2 code). Altivec is a very specialized performance enhancement.
Another - We all know that the Celeron's a pretty crappy CPU, performing far lower than an equivalent speed P4. It's bottlenecked by a slow bus & hamstrung by limited cache - no matter how fast your core is, if the it can't access any data it can't actually do anything. Similarly, if you have two identical P4s, one with an 800MHz FSB and another with a 533MHz bus, for many things you won't see a difference, but in others the 800MHz chip will slaughter the 533.
Ths list goes on and on. All clock speed is is a precisely timed electrical pulse which, by itself, does no computation.
Another - PowerPC chips (like the G4 and G5 in newer Macs) have a kick-ass vector processor with Altivec. G5s running properly tuned software that fully utilizes Altivec can beat or meet a P4 at twice the speed (running good SSE/SSE2 code). Altivec is a very specialized performance enhancement.
Another - We all know that the Celeron's a pretty crappy CPU, performing far lower than an equivalent speed P4. It's bottlenecked by a slow bus & hamstrung by limited cache - no matter how fast your core is, if the it can't access any data it can't actually do anything. Similarly, if you have two identical P4s, one with an 800MHz FSB and another with a 533MHz bus, for many things you won't see a difference, but in others the 800MHz chip will slaughter the 533.
Ths list goes on and on. All clock speed is is a precisely timed electrical pulse which, by itself, does no computation.