Tom's Hardware > Forum > Overclocking > CPUs > what is more important FSB or Multi?

what is more important FSB or Multi?

Forum Overclocking : CPUs - what is more important FSB or Multi?

Tom's Hardware: Over 1.4 million members in 6 different countries available to answer all your high-tech questions. Sign up now! Its free!
Word :    Username :           
 

hi,

let say i have a Q66- 266x9=2400Mhz @ 1066FSB
& the multiplaier can be changed from 6 to 9.

if i can get with a low multi 400x6=2400mhz @ 1600FSB is it going to be faster than 266x9 @ 1066FSB? (because the cpu can get more data with higher fsb?!?!?)

this cpu can get higher i know. so lets say i hit 3.2Ghz with 400x8.

so if i can get it to run 514x7= 3598Mhz ~3.6Ghz @2056FSB is it going to be faster than 400x8 @ 1600FSB?

or it will bring the similar results or the opposite- lower FSB higher multi will bring better results.

thanks.
D.

------------------------------ you don't want to know...
crysis DX10 1280x800 at 35-40FSP is the Target.

Dell 27" 1920x1200 is huge!
Sponsored Links
Register or log in to remove.
- 0 +

The higher the FSB the better. The multiplier is useless other than the fact that it lets you change speed. It does nothing more.

So 400 x 9 = 3600 and so does 450 X 8, but you'll get more data, just like you thought, through the FSB the higher it is. That can especially effect your Memory, because thats what the FSB is for, its the only pathway to you main system memory!

Also, you'll want you memory in sync. Kinda hard to push 1066 Mhz worth of data through a 800 Mhz bus. (400 fsb x 2 for RAM) So again, the higher the FSB, the better the performance!

--Lupi!

Reply to Lupiron

Lupiron wrote :

The higher the FSB the better. The multiplier is useless other than the fact that it lets you change speed. It does nothing more.

So 400 x 9 = 3600 and so does 450 X 8, but you'll get more data, just like you thought, through the FSB the higher it is. That can especially effect your Memory, because thats what the FSB is for, its the only pathway to you main system memory!

Also, you'll want you memory in sync. Kinda hard to push 1066 Mhz worth of data through a 800 Mhz bus. (400 fsb x 2 for RAM) So again, the higher the FSB, the better the performance!

--Lupi!



Intel's Core2's and AMD's A64's dont benifit from a higher fsb anymore 100mhz more on the cpu clock will benifit more then 100mhz on the fsb, but on the other hand call me old fashion, i prefer a higher fsb and sync'd ram.

------------------------------ Q6600@3510/1560 + TT BigTyphoon+Mod
8gb Kingston 800mhz
Gigabyte EP35-DS3P
XFX 8800GT/512
Reply to apache_lives
- 0 +

Aha but would you say that 3GHz on FSB vs. 3GHz on multiplier is better? I would. I think if you're going to OC, may as well raise FSB first.

------------------------------ "He is no fool who gives what he cannot keep to gain what he cannot lose" -- Jim Elliott
Reply to leo2kp

Ok I just bought a E2180 which is FSB 200 (800 QDR) and 10x multiplier for 2.0 GHz total speed. If I overclocked it to FSB 400 (1600 QDR), would that be 1:1 for my DDR-200 4-4-4-12 RAM?

Reply to cmmcnamara
- 0 +

1600FSB or 2000FSB for the same overall Ghz- lets say 3Ghz
is not a 100Mhz difference. it's 400Mhz.
& it will be more sync'd with the DDR2-800Mhz.

so will it preform better?

it's like to compare Q66 266x8 to Q93 333x7.5, but they have some other different specs so it's kinda' hard...

------------------------------ you don't want to know...
crysis DX10 1280x800 at 35-40FSP is the Target.

Dell 27" 1920x1200 is huge!
Reply to z_dori
- 0 +

best to keep your ram in sync with a 1600FSB for DDR2-800 or 2000FSB for DDR2-1000.

The goal for most is to try to max out CPU GHz and FSB while keeping the ram in sync. The multiplier is just something that you can adjust as needed depending on your overclocking strategy.

Edit:

Hmmm.... I see what your getting at, I would say 1600FSB with memory in sync is better choice of the two. However 2000FSB with your memory overclocked to DDR2 1000 would be even better if you can do it.

I'm personally happy with a 1600FSB (because I don't want to raise the voltages on my northbridge) so I run my ddr2 1000 dimms at ddr2 800 with lower timings.


Message edited by pshrk on 05-07-2008 at 11:43:43 AM
------------------------------ Intel E6400 @3.0GHz 1.35vcore, Zalman CNPS9700, eVGA 8800GTS 640MB, Asus P5N32-E SLI, 4G (2x2G) Mushkin Redline DDR2 8000, Thermaltake Armor, Enermax Modu82+ 625W PSU, BenQ FP241W 24" LCD
Reply to pshrk
- 0 +

some one should benchmark it... maybe the difference is not that big...
i think in games the difference will not be so big.
but how about multitasking?!! that is the whole point..
multitasking faster & better...

------------------------------ you don't want to know...
crysis DX10 1280x800 at 35-40FSP is the Target.

Dell 27" 1920x1200 is huge!
Reply to z_dori

This might be as close as you can get. Anand tests memory by changing FSB and multipliers.

http://www.anandtech.com/memory/sh [...] i=2916&p=4

Look at the second chart on this page. Notice that the second line and the fourth line have the CPU at the same speed. (3.29GHz, the memory is the same speed at 1315.) What does the FSB of 411MHz get you over 329MHz? The CPU frequency is the same, and the super pi score shows the same 38.4 seconds. (I believe that Super PI is more a function of CPU "horsepower" and cache then a memory test...) All three of the video games do show a few extra FPS however, even though the CPU frequency is the same.

This should show, at least for Intel systems, that the higher the FSB, the faster your computer can perform. It will of course matter on what test you run, as not all tests will stress the FSB.

------------------------------ The voice of REASON
Do NOT feed the TROLLS!
Always a DEMON!
Reply to 4745454b

The Buss speed used to be very important, not so much anymore.

Toms actually did an article on this some time ago with a ton of benchmarks.
I don't know if a person could find it since they have changed the site a couple of times since then, (It just keeps getting worse, I cannot believe how badly this site is sliding down hill) so I don't know if you could find it or not, I sure couldn't.

By the way, has anyone but me noticed the forums are full questions that anyone who read Tom's would know as common knowlege? I mean over and over and over again. This tells me that very few people actually read Toms anymore. They just jump to the forums to get what they need. Can't say I blame them, I don't pay any attention to what they have on site anymore either. Sorry about the ranting....

Basically it was showing that with buss speeds of 800 and more, raising the mulitplier was better than raising the buss. Since nothing else in the PC actually could run at or benefit from buss speeds any higher. Memory can, but at higher latencies, so any benefit gained from a higher buss speed is negligable.

The higher and higher buss speeds for the moment are just like DDR3, and 1 gig of memory on Graphic cards with a 256 bit memory buss. Just bigger and bigger numbers to entice people into buying new hardware, whether it means anything or not. They have hit the ceiling with clocks speeds on CPU's. They have to keep raising numbers somewhere or people won't be willing to buy new.

Reply to jitpublisher
- 0 +

z_dori wrote :

some one should benchmark it... maybe the difference is not that big...
i think in games the difference will not be so big.
but how about multitasking?!! that is the whole point..
multitasking faster & better...



Been done - by lots of people. You can do it yourself.

For example, 9 X 400 MHz and 8 X 450 MHz both give you a CPU core speed of 3.6 GHz. All other things being equal (and they seldom are), you'd expect, using a good memory benchmark program, that running at 450 MHz would give you a 12% increase in memory bandwidth.

When I did it, I got about a 5% increase. With applications and games, I saw no noticeable improvement. Apparently memory performance does not scale linearly with FSB frequency at the high end. I'm not too surprised about games. I suspect that most modern video cards bottleneck around 3 - 3.2 GHz.

I suspect that's due to two factors. The Intel Core2 architecture is pretty efficient, and the large L2 caches help a lot.

The main thing I noticed was an increase in motherboard temps.

Of course, YMMV.

Message quoted 1 times
Message edited by jsc on 05-07-2008 at 03:01:48 PM
Reply to jsc

z_dori

Back in the distant past when I could actually understand how computers worked (it's the computers that have changed, not me), there were so few factors that affected performance that these kinds of questions could be answered.

I suggest that the correct answer now will almost always be "It depends." What it primarily depends on is which particular application you are using. Not what type, but which application. When I look at benchmarks in recent articles, rig A is fastest on three tests, rig B on four others, and rig Z on three more. Usually not even the same kinds of tests, so you can't say "multiplier is better for gaming and memory for video editing."

I suspect that the only areas where it even matters anymore (!) are gaming, scientific number crunching, video editing, and possible massive audio processing. For anything else, current high-end machines are simply good enough. If it takes me 68 seconds to transcode an album vs 62 seconds, it simply doesn't make much difference to my life.

Things that are now more important to me than raw "speed" are

  • Multitasking, which you mentioned. Multitasking gives me a better perceived response time when I am working and something is running in the background. And something is always running in the background, like an OS or antivirus or firewall.
  • Quiet. My computer sits in my "home office." If I don't have TV or radio on, it's the loudest thing in the room. This despite my having custom-built it with an Antec Sonata case, slow 120mm fans, fanless video cooling, fluid-dynamic bearing disks (which were new when I build it), a honking great Zalman cooler for the CPU, and case damping. It's quiter than any other machine that I've had, but it's a noise source.
  • Fast response. When I open an app, I'd like it to be there. My office machine can take two minutes to open Visual Studio with no project loaded!
  • Fast startup. I've given up on this, what with firewall and antivirus and antispyware and all of these doing updates-on-startup. And the SCSI controller card that slows down the POST process.


I don't think that there are hard-and-fast rules anymore. Even a 1:1 memory bus ratio can be out-benchmarked. The trick seems to be balancing things. For example, Intel used to put out faster and faster CPUs without giving them bigger pipes to the outside world. It got way out of balance, and the CPUs starved for data. Now they are raising bus speeds and lowering multipliers (and saving watts), and it helps, but there is no magic balance that is best for everything. Balance points probably depend on the type of memory, mobo, and cpu you use. Find and eliminate the bottlenecks, and you will be within ten percentage points of the best. Good enough for most.

My personal pet peeve at this time is disk throughput. So I'm playing with a datacenter-grade RAID enclosure which takes up to eight SATA disks, attaches to my PC via Ultra320 SCSI, and sounds like a small plane's engine. I chose the Ultra320 for two reasons. One, higher possible throughput than SATA buses. But really because it can be meters and meters long, so I can have the PC in my room and the plane engine in the nice soundproof attic. I will not use NAS because the bandwidth limitation is below that of a singe directly-attached SATA drive.

Thanks for listening to me, and remember that the key answers are

  • It depends
  • What do you really want the machine to do
  • Comfort factors: noise, heat, fast boot and application startup

Reply to wyomingKnott
- 0 +

thanks jsc!
this is a very good information i was hoping to get.

------------------------------ you don't want to know...
crysis DX10 1280x800 at 35-40FSP is the Target.

Dell 27" 1920x1200 is huge!
Reply to z_dori
- 0 +

Thanks WyomingKnott!

your answer is the general one.
offcourse i hate my pc noise! 12x8cm fans (not including the CPU's & PSU) is a lot of noice. i wish i could hear the Hard Drives... but this is a vacuume cleaner...

the question was from a general thinking about OC....
if i could test it myself i would, but i don't have the CPU or the MB for it...
maybe in a couple of months... with p45 boards.

------------------------------ you don't want to know...
crysis DX10 1280x800 at 35-40FSP is the Target.

Dell 27" 1920x1200 is huge!
Reply to z_dori
- 0 +

Think, peoples! The Multiplier is a fake option. It does nothing by it's self. It needs the FSB to return a speed. How can a multiplier be more important?

Intel does use its multiplier to limit the frequency of its chips. If its locked low, it does make it hard to OC, because, Behold, its all FSB reliant at that point, meaning your Mother boards FSB must crank up to xxx for you to get more speed.

Although its funny to see some arguments, I still have to say that the FSB is a big factor. I am gonna test it right now, have a q6600 at stock with FSB at 266, RAM at 800 Mhz 4-4-3-15, will let ya know! Gonna take a while because I will use super pi's 32 million decimal calculations as my time frame reference.

Let ya know in like 30 mins!

--Lupi

Reply to Lupiron

jsc wrote :


When I did it, I got about a 5% increase. With applications and games, I saw no noticeable improvement. Apparently memory performance does not scale linearly with FSB frequency at the high end.



YMMV? Of course it would, my link showed that. Running the math, the games saw increases of about 2% with the faster FSB. (notice that the ram timings were the same.) The faster FSB does help, the question is by how much?

------------------------------ The voice of REASON
Do NOT feed the TROLLS!
Always a DEMON!
Reply to 4745454b
- 0 +

:: does some quick calculations. :: Uhh, 28 seconds in a 20 min test is what percentage?

 

Anyways, thats the results. I can post the screen shots if you'd all like.

 

While 28 seconds may not seem like much to you and me, rest assured that your computer would appreciate the boost in speed, since its free.

 

(I know I'd rather not wait around 20 mins just to gain a 28 second time advantage, but since you wait for your computer all the time, that would reduce your waiting!)

 

--Lupi


Message edited by Lupiron on 05-07-2008 at 10:07:47 PM
Reply to Lupiron
- 0 +

If I put mine to 400x8 instead of 320x10, would this require me a higher vcore?

Also my current BIOS or board sucks, I can't even adjust the timings for myself only memory multiplier.

Reply to Rocky90
- 0 +

LOL, nice, and no, it would not require any more VCore because its the Board that uses the FSB, the processor would still be at the same speed, and as long as its at the same speed, or a fraction off, you'll be just fine with the VCore you have now.

--Lupi

Reply to Lupiron

The only thing the higher multi dose is it enables lower FSB to reach high clock speeds meaning less stress on RAM (esp. for FSB 1600+).

 

@Rocky90: What motherboard?


Message edited by Shadow703793 on 05-07-2008 at 11:10:02 PM
------------------------------ http://farm3.static.flickr.com/2617/3815217176_0a5be7955d_o.gif
http://farm4.static.flickr.com/3553/3818083596_1a772f7162_o.gif
Reply to Shadow703793
- 0 +

This has been tested time and again, by virtually everone who OCs their computer, including myself.

There is no longer a performance difference between higher multi/lower fsb and lower multi/higher fsb.

None.

Reply to seboj
- 0 +

What???? There is a 28 second time gain in super pi if you dont go with the 1 mil calculations! So obviously thats faster!

Guess i will have to get the pics from my formula, BRB!

--Lupi!

Reply to Lupiron
- 0 +

And the slower one was technically almost 1 Mhz faster! Hahaha!

--Lupi!

Reply to Lupiron

doesn't raising the FSB creates more heat? i'm pretty sure i read it in an article awhile back either on here or anandtech

i guess im a safe overclocker and only went to 400 fsb on my comp.

correct me if im wrong, but i rather go with 400x9 =3.6ghz then 450x8 =3.6ghz.

and as stated before, either you have hella good ram or you gotta OC them to keep up 1:1 with the higher FSB = more voltage = more heat.

------------------------------ http://valid.canardpc.com/cache/banner/590311.png
Reply to aznguy0028
- 0 +

Thats true, other stuff matters! But just plain and simple. The higher the FSB the better the performance at the same speed processor!

Yes, it can effect Mother Board heat, but only minor amounts!

Basically, it does nothing to the CPU at all. Its MoBo orientated, so thats that! if you have a good MoBo made for OCing, the higher FSB is better.

And then your RAM must support it, chipset, etc...

But that wasnt the question was it? The question was which is better.

The answer is obvious for those who wanna do it.

--Lupi!

Reply to Lupiron

Higher FSB does cause more issues for the chipset, but as long as the chipset can handle it, then no worries. It might put out a bit more heat, but I'm not sure how much heat a northbridge can really output.

------------------------------ The voice of REASON
Do NOT feed the TROLLS!
Always a DEMON!
Reply to 4745454b
- 0 +

:) Exactly. If you can do it, and your board can do it, then its free, since you already have the components. Isnt OCing about more speed and power? This is a free way to get a bit more performance!

Enjoy!

--Lupi

Reply to Lupiron
Tom's Hardware > Forum > Overclocking > CPUs > what is more important FSB or Multi?
Go to:

There are 1044 identified and unidentified users. To see the list of identified users, Click here.

Please mind

You are about to answer a thread that has been inactive for more than 6 months.
If you still wish to proceed, please ensure that your posting is original and does not duplicate or overlap any prior responses to this thread.

Add a reply Cancel
Sponsored links
  • Ask the community now
  • Publish
Ad
They won a badge
Join us in greeting them