why aren’t cpus at a native 4ghz yet?

Hello all,

I'd like to know why aren’t cpus at a native speed of 4ghz yet?

I mean come on, since the P4 days, there hasn’t been any jaw dropping RAW native speed increases. Sure there’s multiple core's and drastically improved memory/bandwidth techniques used, but the way i see things panning out, I can understand why nvidia/ati want push past the cpu barrier and take on some of the rolls that of the cpu themselves.

A good example is with the gtx280 and the ati4870 where they're both heavily bottlenecked with today's highest native cpu speeds and is pretty much a requirement to o/c your cpu to the 4ghz region to see any decent improvement compared to the last couple generations of gpu's.

I hope amd's and inte'ls next cpu offerings will change all this, because as it stands now, its nearly a waste of money buying new graphic cards if the cpu just hasn’t got the balls to feed them.

Raw processor frequency is just simply not the way to go anymore. Both AMD and Intel have put their R&D dollars to work on getting more instructions per clock and cramming more processing cores into a processor package.

A single Core 2 core against a single P4 core at the same frequency would be a total bloodbath in favor of Core 2.

AMD was the company that gets the credit for breaking us of our GHZ addiction. They showed us with their Socket A processors and model rating system that their processors would outperform a higher frequency processor. Ever since that time we have relied heavily on benchmarks to tell what processor is better.

Also, as software writers start going multi-threaded (Photoshop, encoding, newer video games like Crysis) we start to see huge gains from multiple cores. It will take a while for software to catch up, but it's happening as we speak!

I beg you to look at performance results, not processor frequencies.


I hope that helps.

We don't all have power plants in our backyard and liquid nitrogen pumped to the house like natural gas. How are we supposed to overclock when the chip is already being pushed?

If you run at 1920x1200 with 4xaa all on max, you aren't going to see much difference going from 3-4GHz. Maybe in FSX, but that is a coding horror.

Go back to your hole you bastage! (But continue to fund my LN2).

While thats generally true, its not an exact truth. Theres several things that make that too general. First off, if youd said or added "for acceptable fps" itd be more true, but even then theres a few things Id like to point out. Using the 4870x2, you cant say that 4xAA has a major effect on that card, as it barely does at all, and many times less than what youd get from a higher oc. Also, once DX10.1/11 comes out, 4xAA will be a moot point, and will give a capable cards a nice fps boost. So far, the only game weve seen it used in, it showed a 20% boost, tho those cards werent very good at AA to nbegin with, and I havnt checked the performance difference on the 4xxx series, which can be found at TR when the 4xxx series was released, as they included the DX10.1 benches on the 4xxx series as well. Needless to say, it made the lower 4850 either beat or come close to the G280, which shows the improvements.
Theres plenty of benches that show the 4870x2 and sometimes the G280 as well still seeing improvements no matter how high the cpu is clocked. In some games, this is important, in most, its not.
One thing a higher oceed cpu will do no matter how good its multithtreaded is in the minimum fps, which is the more important use of fps anyways, so saying higher clocks isnt the way to go, or even higher IPC, isnt true, but mainly due to the law of physics, multithreading is forced upon us, as cpus just cant go much faster, at least until we find a better material to make out cpus from

Its part of the DX model. It allows for fewer passes required, thus adding to or if you will reducing a cycle from the gpu. Remember, its not just a SW requirement, the gpu has to compatable, HW compatable

Fortunately I'll never have a problem, my 9600GT is too slow to need my CPU any faster than 2.7GHz

When your CPU idles at 51C, it's a good thing not needing to overclock

No idea. It's pretty hot in here today, it runs around 45C in winter, which is still warm.

Thats true, and itd be good to see as well, tho Id hate to have to pay for any of those chips
@ da kittah, do you know Satan? heheh thats hot!!!

I feel your pain randomizer.. i got a 8800gt AKIMBO (uber cooling) for my inspirion 530... tight fit ... and it usually under load is about 55 C.. unless i'm playing crysis and that usually takes it up to about 60C, once it even got to 65C.. and thats with the fan on 75%

Needless to say, im looking for a new case... lol unfortunatly, i haven't seen one so far to slat my lust though i do have about 10 bookmarked on newegg eheheh.

hah don't feel so bad, coretemp 0.99.3 reports my

Q9450 @ 3.2ghz to be 46,49,45,46

while RealTemp 2.70 reports 41,41,44,41

and this is at night at like 2AM, during the daytime, it hit 50's idle easy. priming takes it up to around 65-67 and this is in a clean wired antec 900 case with front/exhaust/side fans on high. every fan is 120mm, and the top is 200mm which is on medium.

funny how intel in the P4 days thought they could get up to 10ghz with the netburst architecture.

they encountered a little thing called heat

I ran a series of tests with my E8400 at 3.6Ghz, and at 1.2Ghz... there was barely any bottlenecking. In fact, I still got about 80FPS (vs my usual 120) in Call of Duty 4 with a timedemo (1680x1050, Max, 4xAA, 8xAF). I think anything E6600 or above should be plenty to handle any single card/chip solution.

yeah, in 5 years we'll have 64 core processors, still clocked at 3.0Ghz for an extreme edition, and the caches will still be like 3MB per core ... seriously though.

It seems like the computer companies these days are just learning to shove more of the same stuff into the chip casing.

Sorry for the horrendous amount of posts lol.

five years??? Try 2-3 tops. Next year/early 2010 Intel is releasing a quad core with octo-hyper-threading.

J believe that's 32 instructions per cycle, or the equivalence of 32 cores. I'd say 2-3 tops. (5 years might be how long it will take the gaming industry to shift to the idea of large number of CPU cores, which will be awesome for games. It'll be a nightmare for those companies that are refusing to modernize their coding base for multi-threaded languages, though.)

So many smilies... Since when did spam bots become moderators?

*snicker*

*Holds up anti-ban shield*

Am I missing the point here but isnt it how many clock cycles a processor has or even mips that makes the difference..

Yes at one point MHz was the race between Intel and AMD...

AMD proved that wrong because it could do more per clock count..

Intel did this with core 2 duo... And continue to do so - so very much better than AMD..


It makes me laugh when i see on ebay a 10GHZ Q6600 processor when there is no such thing
each core runs a 2.4 but its what each core does with that 2.4 that counts..

Intel proved that with the Centrino laptop chip at 1.7 comparing to p4 3.4s

I would rather have a 2GHz running at low power with a better efficiency rating which has more raw power than a 4GHz anyday.

a 2.4 is the new 4GHz in p4 terms...

Now obviously the more MHz longterm is better but this means more heat which has other problems to contend with..

Problem is, the more efficient ( fewer cycles, shorter the pipe) for a better IPC, the hotter it gets, or the slower it goes. So, finding a good balance , or sweet spot is the key, and since C2D, Intels pretty much found it, and just has needed smaller processes and a afew tweaks to reach the low 3Ghz range. Someday, MAYBE a 4Ghz could come, but well have to wait for awhile on that

I have a Centurion 5. Personally, I think the cooling sucks. I have to Scythe Ultra Kaze 3k RPM fans....one intake and one on cpu cooler and a diff brand 120mm for exhaust. My cpu idles at 42C when my room temp is 75 +- 2degrees. maybe when everything cools off in a few months I'll be seeing decent temps.

I forgot to mention. My case is on it's last leg before i throw it outside to feed the garbage dumpster.

Since Forum Kiitahs wished for LN2 piped to their houses....

A 4GHz Yorkfield is possible at around 150W (probably less with the E0 stepping, the following numbers are from the C0 stepping).

http://www.anandtech.com/cpuchipse [...] i=3184&p=2


Its not as far fetched as some people think, but there is simply no need for Intel to compete with itself since the top AMD quad is clocked at 2.6GHz and is slower per clock as well.