Han kan inte vara kunnig om han tror att hög mhz är allt. Läs eng text nedan, taget från länken.
http://www.cooltechzone.com/index.php?option=content&task=vie...
Differencing Between Myths and Facts?
Most, if not all, of you have read or been in heated debates over why or why not higher clocked microprocessors are better than their "comparable" competition. Just why is it that so many people stand behind their arguments about a 2.4GHz processor outperforming a 3.4GHz processor overall? Also, why do benchmarks sometimes show lower clocked processors beating higher clocked processors in some tests? I thought gigahertz and megahertz determined a processor’s performance… Well, you’d be surprised if you were the one who thought that.
There has been fierce competition between the two main leaders in the microprocessor market for several years now. It used to be that Intel was typically the hands-down choice for the best performance, but has that changed now? Well, the scale seems to be tipping in favor of Advanced Micro Devices (AMD). However, there have been similar tips of the scale in the past. First it seemed as if Intel was undoubtedly on top with the Pentium I and II microprocessors, but things changed. The K6 microprocessor was probably the start of AMD’s advancement in the microprocessor industry. Although the K6 fell short of being the overall performance leader, it laid the groundwork. As the new millennium neared, the scale tipped back and forth, keeping both corporations on edge. The Pentium III processors were quickly becoming popular, but AMD’s Athlon processors gathered much recognition as well.
Just before the dawn of the 21st century, Intel announced its release of the Pentium 4 processor. The initial processors were clocked at 1.4 and 1.5GHz. Similarly clocked Athlons still performed competitively with the Pentium 4 line, but it was obvious that Intel clearly didn’t want AMD to interfere with the scales. The Pentium 4 line grew as 2001 came along and Intel ramped up the processor clocks leaving AMD behind at 1.4GHz. AMD seemed to focus more on the value market as it released the 1.0GHz Duron processor instead of pursuing the quickly accelerating Pentium 4 processors which hit 2.0GHz in August of 2001. This wasn’t entirely true, as the 1.4GHz Athlon processor excelled in several benchmarks over the higher clocked Pentium 4 1.8GHz CPUs. With a 400MHz difference, it is obvious that the clock (CPU frequency) doesn’t always determine whether or not a microprocessor will perform better compared to another. This became even more noticeable as gigahertz differences between processors still showed comparable results in performance with the Athlon XP and the Pentium 4 CPUs.
Differencing Between Myths and Facts? Continued
Realizing the negative image normal consumers would receive from seeing lower clocked processors, AMD switched the nomenclature system of its microprocessors to more of a comparison rating. Thus, 1.67GHz Athlon XP "2000+" processors were to perform similarly to 2.0GHz Pentium 4s, overall. As it turned out, Athlon XP processors of even lower ratings (e.g.: "2500+" vs. 2.8GHz P4) outperformed the competition in many benchmarks.
In some benchmarks, applications favored Intel’s Pentium 4 architecture and thus were noticeably superior to comparable AMD products. These applications were mainly related to multimedia and video and sound production/editing. Intel still holds the reins with these benchmarks. It is generally agreed that both companies offer excellent products for their intended application (e.g.: value, performance, etc.), but many still have confusion over why higher clocked processors don’t always perform as well as their lower clocked competitors.
For example, processors may have similar clock rates but one may perform better over the other. The reason is due to the architecture of the microprocessor, which thus determines the clock cycle (rated in megahertz and gigahertz) and the instructions per clock (IPC) that the CPU is able to perform. CPUs with longer pipelines tend to have higher frequencies than their shorter-pipelined counterparts. This is not always a good thing, as the longer the pipeline, the longer it takes for a single instruction to be processed.
As a result, the shorter pipelines in the Athlon XP architecture, for example, results in a lower clock but an overall higher amount of instructions per clock. Now you may be wondering why the number of clock cycles is more heavily looked upon than the number of instructions per clock. For one, IPC only tells you how many instructions per clock the processor can perform and it doesn’t really tell you anything about its performance. The "GHz" and "MHz" ratings do give an overall idea of how well the processor will perform, but there needs to be a general variable that can give someone an idea of a generalized performance rating.
Differencing Between Myths and Facts? This time with examples...
In our example performance rating (PR) model, you can get an idea of what you can truly expect, performance-wise, from a processor by combining two important factors—clock cycles and IPC.
Athlon XP: XP2400+: 9 IPC 2000MHz
Intel P4: 2.4GHz: 6 IPC 2400MHz
If you multiply IPC by clock rate, you can arrive at a generalized PR and conclude that the Athlon XP 2400+ has a PR of 18,000 (also number of instructions executed per total-cycle) compared to the 14,400 PR of the Intel Pentium 4 2.4GHz processor. As a result, you can expect the Athlon XP 2400+ models to perform somewhat better in general usage. However, media applications have tended to favor the Pentium’s longer-pipelined architecture and thus, the PR rating cannot be effectively used to reach an appropriate decision.
Another flaw of our example — which is only used for generalized purposes — is that it doesn’t take into account the possibility of differing cores with different architectures, and thus, you can expect a noticeable difference in the performance rating of each.
You must never look only at the combined GHz and MHz ratings of a particular microprocessor, but instead, should compare the number of clock cycles per second (GHz/MHz) with the IPC and architecture in order to make a decision of purchase. As of now, you have several types of modern processors to choose from: Pentium 4, Celeron, Celeron D, Pentium M, Athlon XP, Athlon 64, Athlon FX, and Sempron. The Intel Pentiums (with the exception of Pentium M) and Celerons tend to have more stages in their pipelines than the Athlons and Semprons, thus depending more on their clock speed to "keep up" with lower clocked Athlons and Semprons. The Pentium M has a similar amount of pipeline stages as the Athlons and Semprons. The decision on which processor to use will probably be founded upon the main application you intend to use it for. Just remember, gigahertz and megahertz aren’t everything!
Skrivit lite själv på ett lättare sätt så han kanske kan förstå det:
Lite om PR-rating och hur effektiv en CPU är.
- MHz är inte det enda som avgör hur snabb en processor är utan bara en av många variabler som spelar in. Det är bla en kombination av MHz och hur mycket arbete som görs varje klockcykel(IPC= Instructions per cykel).
- Prestanda = Antal MHz * Arbete.
- En processorer gör antingen mycket arbete få gånger per cykel, eller mindre arbete fler gånger per cykel.
- AMD's PR-märkning kan man säga är ett riktmärke på hur snabb en CPU är. Det är ett ungefärligt riktmärke på hur de skulle se ut om både Intel och AMD skulle ha göra samma mängd arbete per klockcykel. Gäller iofs samma CPU-generation, se bara XP VS P4. XP är bättre än P4A, likvärdig med P4B men mycket sämre än P4C. Däremot är AMD64 bättre än alla P4 utom P4EE(Extrem Edition).
- PR-systemet är till för att vanliga människor snabbt och enkelt ska kunna jämföra AMD och Intel på ett relativt rättvisst och korrekt sätt.
Slutsats:
- Intel har en hög klockfrekvens men ineffektiv sett till hur mycket arbete den gör per cykel.
- AMD har en lägre klockfrekvens men effektivare sett till hur mycket arbete den gör per cykel.
- Sen så finns det mycket mera som spelar in, tex längden på pipeline, cache-minnets storlek, FSB, osv. Tippar att jag inte ens vet hälften av allt men detta är iaf lite om varför PR-rating finns samt att MHz inte är allt.
