Sign in with
Sign up | Sign in
Your question

e4300 e4400 e6320 e6420: which one to choose?

Last response: in CPUs
Share

Which do you want?

Total: 183 votes (17 blank votes)

  • e4300
  • 15 %
  • e4400
  • 29 %
  • e6320
  • 17 %
  • e6420
  • 41 %
February 22, 2007 9:55:00 PM

In April the e4300 drops in price and 3 new models come out, but which one to choose?

This discussion assumes that all chips will be Over Clocked

e4300
FSB: 200mhz
Multi: 9
Cache: 2m
$113

e4400
FSB: 200mhz
Multi: 10
Cache: 2m
$133

e6320
FSB: 266mhz
Multi: 7
Cache: 4m
$163

e6420
FSB: 266mhz
Multi: 8
Cache: 4m
$183

1. Is the extra cache on the e6x20 enough to make up for the lower multiplier? For the lower multiplier AND the higher price?

2. Which chip will have the most FSB bandwidth after OCed and how much does it matter?

3. Is the 10x multiplier on the e4400 too high? Will it run too hot? Will it limit FSB enough to hurt performance? Will it limit the OC at all? Is it worth the extra cost?

4. Which chip will get attain the highest internal frequency when overclocked?

5. Do you care or will you buy one of the <$100 competing CPUs? Or perhaps a more expensive CPU? If so, which one? Or will you wait for Agena and choose then?

-------------
My intial responses:
1. No
2. e6320, no clue if it matters though
3. 10x multiplier is not too high to limit the OC as long as you have enough cooling, not sure if it's worth the extra cost and heat/cooling though
4. e4400
5. Brisbanes are temptingly cheap, and if OCed to +2.4ghz I will probably be perfectly happy with the speed, but I can't decide. Maybe I should wait until Agena, but I have a dying system in my isntall base that needs replaced soonish. Help?
February 22, 2007 9:58:24 PM

E4300

9*400=3.6ghz 2mb cache... Looks good to me. If it can hold that speed, im all for it.

10*333 = 3.33ghz on a FSB limited baord would be good. Save some $$ on ram and MB.
February 22, 2007 10:01:41 PM

none - the e6600 is the better - the faster bus just means you have to spend more for bigger multiplier.

those new chips are for dell etc - smart people will stick with e6300-e6600 the x6800 for us with savings to waste!

I voted for the e4400 it the best of the 4 for cheap fast system!

IFB (AMTI is always #2) The answer is C!
Related resources
February 22, 2007 10:15:24 PM

e6600 has same multiplier as e4300, all you get is a bigger cache and it costs twice as much. e4400 has a bigger multiplier and once both chips are OCed the e4400 should win at at least some benchmarks and shouldn't trail by much at anything else. If you want something that is all-round better than the e4400 the e6700 is the next cheapest thing you should be considering. I wouldn't buy an e6600 even if I desperatly wanted to spend more on a CPU. For the $100 you would save choosing an e4x00 over an e6600 you could buy a very nice HSF and have a few $$$ left over to spend elsewhere, if you truely have no budget e6600 is no place to stop xD

Seriously, "smart" people shouldn't even be considering e6300 or e6400 now that e4300 is out, and once e4400 comes out... it just doesn't make sense when there are cheaper chips that have more potential. Plz prove me wrong if I'm crazy but the higher multiplier and Allendale core are simply better than the Conroe e6300 and e6400. Not sure why it's taken people so long to catch on to this, it's been discussed for months already. Here's an article from November 19th 2006. If this isn't an indictation of blind intel fanboism I don't know what is: They come out with a BETTER processor that is also cheaper and people still reccomend e6300 like crazy *shakes head*. wooo Conroe advertising machine! Seriously, Allendale is where it's at right now, no?
February 23, 2007 4:37:55 AM

Your paying for the cache the large "cash" takes up alot of space that cost alot of money hence the price difference - i was under the impression the 4mb vs 2mb made a big difference????????

we need some good data on cache on multitasking performance - i have seen next to none - only claims by the chip makers that the cache makes a big difference.

Why else would spend so much money - again, its a huge amout of chip relestate to go from 2mb to 4mb
February 23, 2007 4:53:39 PM

Thanks for replying jack. It looks like you're trying to avoid making too many guesses as there aren't any benchmark comparisons that address some of the questions I've aske but please play along ;)  I promise not to sick BM on you if you turn out to be wrong :p  The processors aren't out yet, so guesses are exactly what I'm looking for at this point.
Quote:


1. Is the extra cache on the e6x20 enough to make up for the lower multiplier? For the lower multiplier AND the higher price?

The data that has been collected shows that the extra 2 meg cache is not a huge factor, performance ranging from 3-8 % on average with some apps seeing 10 or 15% gains, but few.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=279...

good info, ty. Average difference for the games that they benched was only 2.8% faster with the increased cache at the same frequency. Do you think an additional 28% internal frequency (the difference between e6320 and e4300 at any given FSB) would make up for this difference or would you need a faster FSB clock to make up for the smaller cache? Hypothetically in terms of raw performance I think faster FSB and RAM is the only thing that would make up for the smaller cache, but at the application level I dunno.

Quote:


2. Which chip will have the most FSB bandwidth after OCed and how much does it matter?

The FSB BW scales with system clock speed, so if you OC to a system clock of 300 MHz (your CPU speed would be 300Xmultiplier) the FSB for any processor would be 1200 MHz, at 64 bit wide this would come to 9.6 GB/second, regardless of the process you used. I.e. An E4300 @ 300 MHz system clock would have the same FSB BW available to it as an E4400 @ 300 MHz.

Overclockers know that, but with different multipliers and cache the different chips could hit heat and internal frequency barriors before hitting a FSB wall. The question then is which, if any, of these chips have lower potential for FSB speed and how much does it matter? Would it be better to have an e4300 @ 300mhz FSB (2.7ghz) or an e4400 at 271mhz FSB (2.71ghz) if that was the hypothetical OC limit for each chip all other things being held equal? Of course, the answer to this question isn't very relevant if this hypothetical situation turns out to have no real-world equivalent... With internal frequency differences as high as 43% (e6320 vs e4400 at any given FSB) I think it likely that one will hit an internal/heat barrior first while the other hits a FSB barrior, but the additional cache could equalize them. Thoughts?

Quote:


3. Is the 10x multiplier on the e4400 too high? Will it run too hot? Will it limit FSB enough to hurt performance? Will it limit the OC at all? Is it worth the extra cost?

It depends on exactly how agressive you want to overclock, for the low multipliers the OC is hitting FSB walls before the CPU runs out of gas, if you are really into overclocking then the 10x multiplier and boards with ulocked downward multiplier settings gives you much more flexibility, I would choose the 10x multiplier option for this reason alone.

A very good point, but it would kinda suck to pay extra for a 10x multiplier and then set it to 9x. With data from other Conroe chips with 10x or higher (e6700, x6700, x6800, and mobile C2D chip if anyone has bothered to OC them with better cooling), multipliers we should be able to hazard a guess. e4300 vs. e4400 should be roughly equivalent to e6600 vs. e6700. This really only applies to extreme OCing though, 10x multiplier should be just fine for at least ~300mhz FSB with air cooling and small voltage bumps I think.

Quote:


4. Which chip will get attain the highest internal frequency when overclocked?

Generally speaking, the smaller cache chips will likely give you more % over stock overclock capabilities as there are fewer transistors and less heat that must be dissipated, makes cooling slightly easier.

Easier enough with smaller cache and higher multi to make up for the smaller cache (ignoring price differences for the moment)? Sounds like you're saying e4400 has a +3 multi advantage and a cache advantage over the e6320. Given how many people have voted for e6320 it might be helpful if you stated this explicitly. Until there is real-world data that says otherwise I'm thinking we can expect e4400 to be the king of the sub e64xx world and will be competetive with the e6600 also.

Quote:


5. Do you care or will you buy one of the <$100 competing CPUs? Or perhaps a more expensive CPU? If so, which one? Or will you wait for Agena and choose then?

There is not doubt that Agena will be a big improvement over K8 in terms of IPC, the question here is clock speed --- AMD's 65 nm has shown us less than stellar performance, with clock ceiling limits well below that of the 90 nm windsor cores. It is a gamble in my opinion for a few reasons, after watching out for more data as it becomes available, I was initially wrong that AMD left eSiGe out, they have indeed integrated this material. As such, a top bin of 2.6 GHz and clock limits of 2.8-2.9 (occasional 3.0) OCs, I am not expecting much more improvement from their 65 nm process --- add to this the fact that their new CTO has stated that '65 nm is old news, all focus is on 45 nm' and it is unlikely that Agena will clock up much past 2.8 or 2.9 GHz (in my opinion).

Unfortunately, commiting to one or the other sorta locks you there so the real decision is yours.... we already have great data on the OC potential of Conroe, and Penryn will do even better most likely as the power limitations are going to be removed for the most part. However, there is no guarantee that Penryn will drop in either so this is also a gamble.


My 2 cents anyway

Jack

More info on Penryn and Agena compatibility and prices would be nice, but without any info to suggest otherwise I don't think there are any events on the near horizon as significant as this e4300 price drop and e4400 introduction. These chips just scream "overclock me".
February 23, 2007 5:34:10 PM

Quote:
Your paying for the cache the large "cash" takes up alot of space that cost alot of money hence the price difference - i was under the impression the 4mb vs 2mb made a big difference????????

we need some good data on cache on multitasking performance - i have seen next to none - only claims by the chip makers that the cache makes a big difference.

Why else would spend so much money - again, its a huge amout of chip relestate to go from 2mb to 4mb


Eh, we don't care how much INTEL spends on the chip, we care how much we spend and how much performance increase it gives.

In the link that Jack just gave I believe teh Internet Content Creation (second benchmark in the list) is a multi-tasking benchmark with a 10% performance increase from extra cache at the same speed.
THG's CPU Charts have two multi-tasking benchmarks and they have the e6400 allendale (2m 2.13ghz) and the e6600 conroe (4m 2.4ghz) in the list. The e6600 outperforms the e6400 by 5 and 16% respectively on the two benchmarks but it also has a 12% internal clock advantage so the performace increase from the extra cache is less than 5-16% which pretty much falls in line with the other single-tasking benchmarks. Theoretically the extra cache should help even more when mulit-tasking, but this also depends on Intel's shared cache architecture being highly effecient as each core doesn't have a large dedicated datacache like AMD's processors do. It might not be all that effecient which would explain why the extra cache doesn't give additional benefits when you would think it should, I don't think anyone has really put it to the test since the Conroe architecture was so far ahead of the competition in other places. Ineffecient shared cache architecture might be where AMD is getting their "40% faster" number on select benchmarks for Barcelona (which has dedicated L1 and L2 cache for each core with a shared L3).

But to get back on topic and summurize:
0-15% performance increase from the extra cache depending on the application, single or multi tasking. 15%, I think, would be a "big" difference, but there are very few applications that really benefit. For a multi-purpose desktop machine I would say 6% average maximum faster when running applications that can actually benefit from the extra cache (a generous number I got when averaging all the benchmarks at anandtech and THG that had at least 2% performance increase). I would agree with the anandtech article that larger cache will become more important in the future, but I don't really recommend spending extra money now to "future proof" your system (paying at least some heed to your upgrade path is prudent but you're ussually better off spending extra money later rather than sooner).

Wether or not 6% is "a lot" is a little subjective and depends on what you're doing, and the extra cache may limit your OC and the lower multiplier will almost definately limit your OC, combine those two negatives with a higher price and I don't think e6320 or e6420 are attractive options. If you wanted to go for an e6600, e6700, or x6800 I'd say it's not worth the money but they do have a high enough multiplier and extra cache and should be faster once OCed (with the e6600 vs. e4400 being iffy).

At this point it's still largely conjecture, but from all the data available an OCed e4400 is going to be hard to beat and should require at least an e6600.

(ok, so I'm bad at summarizing, shoot me)
February 26, 2007 6:18:24 PM

I heard a mention of e6320 and e6420 being specced to run at lower voltage but I can't find any articles to confirm it. Has anyone else heard this? this table shows all 4 of the new processors using the same voltage range.

Also, what are the differences between the Allendale, Conroe 2m (e6300 e6400), and the Conroe 4m achictecure used in the new e6320/e6420 vs. Conroe 4m in existing products such as the e6600 if any?
February 26, 2007 8:46:09 PM

There are pin mods to make E6x00 1333 native (assuming the motherboard supports it) and E4300 1066 native.

This can help with FSB limited motherboards. Native 333 doesnt stain the NB and some people on xtremesystems had boosts in their overclocks from this.

Thought I would throw in a wild card.
February 26, 2007 9:28:29 PM

Quote:
There are pin mods to make E6x00 1333 native (assuming the motherboard supports it) and E4300 1066 native.

This can help with FSB limited motherboards. Native 333 doesnt stain the NB and some people on xtremesystems had boosts in their overclocks from this.

Thought I would throw in a wild card.


I've seen your sig before and I wanted to ask: did you really let that Aspire PSU take out all that hardware or did something else fry it?

I don't know why making the motherboard set the CPU higher automatically would make it run any faster unless you were just shortcutting around motherboard limitations... a 333bus is a 333bus. I would asume that any precieved increase in maximum stable OC was due to some other factor.
February 26, 2007 10:27:04 PM

just an honest question. i was lookin at the e4300 and i saw this review in anandtech. they had overclocked it to something like 3.3ghz yet it couldnt beat an e6700 at 2.93ghz. seeing it perform closely to the e6600 i was wondering if the cache does make a difference according to this review?

http://anandtech.com/cpuchipsets/showdoc.aspx?i=2903&p=...
February 27, 2007 1:52:53 AM

Quote:
There are pin mods to make E6x00 1333 native (assuming the motherboard supports it) and E4300 1066 native.

This can help with FSB limited motherboards. Native 333 doesnt stain the NB and some people on xtremesystems had boosts in their overclocks from this.

Thought I would throw in a wild card.


I've seen your sig before and I wanted to ask: did you really let that Aspire PSU take out all that hardware or did something else fry it?

I don't know why making the motherboard set the CPU higher automatically would make it run any faster unless you were just shortcutting around motherboard limitations... a 333bus is a 333bus. I would asume that any precieved increase in maximum stable OC was due to some other factor.

The Aspire wasnt at fault. The metal backplate on my TT BigTyphoon wasnt isolated properly (my error. Shhh, dont tell Intell/Gigabyte) and it came in contact with a few voltage regulators and other solderjoints, shorting the MB, CPU, and doing in my PSU.

It almost happened on my replacement motherboard (965G From Frys), and I realised my mistake after close inspection.


As for the topic at hand, I dont know the exact details, and its fairly recent, so there are few people that have done it and most of them experenced a benifit. Usually it was small though. But hey, if it works, it works 8)
February 27, 2007 4:31:05 AM

Hey Flasher -

Thanks for the PM. Apparently I posted my question in the wrong forum. Such is life. Anyway, so after reading this thread am I to understand that the answer to the question of whether or not the larger cache of the E6*20's has much of an impact on real world performance is sort of an unknown? Or put more simply, per the info thus far, the performance increase is negligible enough that it really doesn't matter on a modestly OC'd general purpose rig?

I think I am pretty set on the E4400 and I ultimately would be happy with 3ghz but one thing I still am confused about is OC'ing and heat issues. If say an E6320 and E4400 are OC'd to relatively the same freq, say around 3ghz, and all other things are relatively equal which will produce more heat? From what I have read I think the answer is the E6320 because of the higher cache (more on the chip). Is this even remotely correct?

.
February 27, 2007 5:32:28 AM

well, the e4 doesn't have hardware virtualization support...
February 27, 2007 5:14:43 PM

Quote:
Hey Flasher -

Thanks for the PM. Apparently I posted my question in the wrong forum. Such is life. Anyway, so after reading this thread am I to understand that the answer to the question of whether or not the larger cache of the E6*20's has much of an impact on real world performance is sort of an unknown? Or put more simply, per the info thus far, the performance increase is negligible enough that it really doesn't matter on a modestly OC'd general purpose rig?

I think I am pretty set on the E4400 and I ultimately would be happy with 3ghz but one thing I still am confused about is OC'ing and heat issues. If say an E6320 and E4400 are OC'd to relatively the same freq, say around 3ghz, and all other things are relatively equal which will produce more heat? From what I have read I think the answer is the E6320 because of the higher cache (more on the chip). Is this even remotely correct?

.


We know the impact the cache has on real-world performance. About 0-15% depending on the task clock-for-clock. The question that is unknown is what will the difference be after both chips are clocked up to a given stability, heat, FSB, or voltage threshold. The e4400 should obtain much higher core frequency all other factors being held equal, possibly slightly lower FSB. At this point a theoretical application that requires lots of FSB and memory bandwidth, low computational power, and a datacache >2m would run better on the e6320. As with the various benchmarks linked above though many applications don't really fit that bill. The benchmark that brick88 linked gives a rough idea of the answer to your question. 2% more FPS in a game is neglible. Quake4 gets 13% more FPS and that's pretty decent but it'll cost more also. Unfortunately in this test the e4300 was running at 375mhz FSB with a 9x multi and the x6800 was running at 266mhz FSB with an 11x multi and this is a good bit different from what an OCed e4400 vs. OCed e6320 will look like. If you managed to get both chips to 375mhz FSB the e4400 would probably beat the e6320 as it would have a 1.125ghz internal clock advantage and it would be closer to the OCed e4300 vs. stock e6600 you see in the graph.

IMHO with the data currently available
Q: Does the cache make a big difference?
A: If you want to call 0-15% a big difference, sure. I won't argue with you. But I wouldn't call a 2% difference big so it really depends on what you're doing.

Q: Big enough difference to make up for the extra cost?
A: Depends on what you're doing and your budget. A personal choice, if you will.

Q: Big enough to make up for a lower multiplier when overclocking?
A: Absolutely not. e4300 and e4400 are both > e6420. If you want faster get an e6600 or better.
February 27, 2007 5:15:47 PM

Quote:
well, the e4 doesn't have hardware virtualization support...

what do you need hardware virtualization support for?


*edit*
according to intel the e6300 and e6400 do have VT and they all have EIST, I looked at a bad link :oops: 
February 27, 2007 5:39:28 PM

I'm kind of surprised by how many votes the e6420 has gotten, I wish more people would post why. An 8x multi isn't so hot when you could choose a 10x multi for a lower price.

I'm also surprised by how few the e4300 has gotten. At $113 tray I thought they would get more lovin. Considering how popular the e6300 and e6400 were and they didn't have 4m cache either. The e4300 is better and will be cheaper than e6400 or e6300 ever were and they have been reportedly OCing quite well.

Is it just name recognician with the model numbers? Do people not trust the allendale core to OC well yet? Is there some kind of price threshold near $180 where the voters just don't care if it's cheaper anymore?
February 27, 2007 7:10:44 PM

Me too.

I would still like to know why the E4300 @3.3 isnt faster than the x6800 (according to the link above)... Its only a 10% clock difference and a 28% FSB difference. I would think thats enough to offset the cache difference.
February 27, 2007 8:18:34 PM

Quote:
Me too.

I would still like to know why the E4300 @3.3 isnt faster than the x6800 (according to the link above)... Its only a 10% clock difference and a 28% FSB difference. I would think thats enough to offset the cache difference.


In 2 of the 3 games tested in that roundup it basically IS enough to offset the difference in cache. 2-3% difference in FPS is neglible. It really depends on the application. Some applications will really benefit from that extra cache, some won't benefit at all.

So early indictation that a ~2.9ghz e6x20 beats a ~3.3ghz e4x00 by 2-13%? in games? Here's a little table of the FSB you would have to hit on each of the 4 processors to hit these frequencies:
CPU - FSB - (OC difficulty)
e4300 - 367mhz - (medium)
e4400 - 330mhz - (easy)
e6320 - 414mhz - (hard)
e6420 - 363mhz - (medium)

Taking OC difficulty and cost into account I think perhaps I am pursuaded to think that for games e4400 and e6420 tie for first, e4300 takes second, and e6320 looses (as do e6300 and e6400 of course).

This suggests that at ~13% slower clock the 4m cache can achieve ~2-13% more performance in games which is certainly compelling. But the e4400 has a 25% higher multiplier than the e6420. FSB will probably be the major limiting factor in OCing these chips wether it's the mobo, ram or the chip itself so I'm not so sure that comparing a 330mhz FSB e4400 to a 363mhz FSB e6420 is realistic. If you were to compare them both at teh same FSB speed the performance advandage of the e6420 (ex: @2904mhz) over the e4400(ex: 3630mhz) would then be something like -7% (yes, negative) to +3%, no? That's a very small difference and the e4400 is $50 cheaper wich you could either take to the bank or spend on more ram/GFX. Or would the e4400 be an order of magnitude harder to cool and thusly not directly comparable? Discuss.

I'm probably still going to buy e4300 though xD $70 difference between e4300 and e6420 pays for a gig of ram :) 
February 28, 2007 3:37:56 AM

Quote:
I'm probably still going to buy e4300 though xD $70 difference between e4300 and e6420 pays for a gig of ram


What I was thinking too. The difference between a 1950xt and a 8800gt (384mb?)
March 10, 2007 11:14:39 PM

Does anyone have any idea why the E4400 is being advertised as "32 bit"? Isn't it supposed to be a 64 bit proc? Am I missing something?

I froogled "core 2 duo E4400" and every e-seller that is taking pre-orders has "32 bit" in the description.

Here is one example. What does that imply exactly? Anyone know?
March 11, 2007 12:28:48 AM

um... the seller is dumb? That's my best guess.
March 12, 2007 1:58:28 AM

Quote:
um... the seller is dumb? That's my best guess.


Well, perhaps but given that nearly every e-seller is listing it as 32 bit I think there might be something more to it. Like maybe that is the description they got from Intel...which still doesn't answer my question though. Maybe I will try emailing Intel.
March 12, 2007 4:20:03 AM

Intel would not tell you anything. The E4400 is still an unannounced product.

Go to Intel's Processor finder site. Click on the E4300 and read off what features it has. The E4400 will have the same just a higher multiplier.

Processor Finder
March 12, 2007 1:39:22 PM

Quote:
just an honest question. i was lookin at the e4300 and i saw this review in anandtech. they had overclocked it to something like 3.3ghz yet it couldnt beat an e6700 at 2.93ghz. seeing it perform closely to the e6600 i was wondering if the cache does make a difference according to this review?

http://anandtech.com/cpuchipsets/showdoc.aspx?i=2903&p=...


per those benchies... the e4300 destroys everything in a price for performance ratio. Only a 10% difference from an X6800 @ stock... com'mon guys. 8O

e4300 gets my vote for 2007. Then get a cheap quad in early 2008.
April 7, 2007 4:41:38 AM

It looks like the E4400 may now be available. I have not yet confirmed it and other stores still have it listed on "backorder" but this e-seller says they have them available. If true I am not sure how NewEgg can justify keeping the E4300 price so inflated when the E4400 is going for $10 to $15 less (sans shipping considerations).

Er... perhaps this warrants a new topic.
.
April 11, 2007 12:33:50 AM

I always thought having a lower multiplier was better for overclocking.

I know I'm new and that probably sounds stupid, but check this link out:
http://www.thetechrepository.com/showthread.php?t=30

Basically, it says that lots of fsb walls are misdiagnosed - they are actually the chipset hitting its limit. It goes on to say that having a lower multiplier (original multiplier, not down shifted multiplier) doesn't stress the northbridge so much in overclocking.

If I am completely off track, can someone please explain to me why you would want a higher multiplier?
April 11, 2007 1:03:35 AM

Eh, that's really not what that article said at all. It says that lowering multi below stock multi can induce FSB barriers earlier on and increase latencies lowering performance. That's all.


Higher multiplier is generally better (in teh case of the e4300 and e4400 it's certainly fantastic), but the allendale core doesn't appear to have as much OCing room in it as the Conroe core does. If you are content with a mere 2.4 -3.3ghz C2D obtainable with value RAM this is "lack" of overclockability is a non-issue. Once you increased your FSB past the stock frequency of the fastest C2Ds currently available you'll be hard pressed to find an application that will suffer from a "lack" of FSB or Memory bandwidth either and if you're ok with that you can just keep the savings in money, electricity, and heat and all you really miss out on is a very, very minor amount of bragging rights for hitting a high FSB frequency.

Heh, I already bought my mobo but if I had known AMD was going to drop it's prices so significantly and could find a uATX AM2 mobo with optical SPDIF I wouldn't even be thinking about intel right now (x2 3600+ is HALF the price of e4300. $60... That's a gig of ram, or a significant upgrade if applied to GFX card).
April 11, 2007 1:07:35 AM

E6320 for me. Cheapest option with the full 4mb L2 cache. That can help improve performance by about 5% when overclocked.
April 11, 2007 1:17:10 AM

Wow,I'm not even an Intel man and Like a blind man searchin for a tuna factory 8) I say the e6600 is goin to be the best bang for the buck, by the time everything is released anyway.
April 11, 2007 2:38:15 AM

Quote:
Quite simple, for overclocking, you want the e4400, because it has a very high 10 mult like the e6700 meaning wiht some nice cooling you could easily hit 4ghz (recommend water for stable, and this is in theory, no guarantees by the taco). While the e63/420 chips are a waste of money imo, because the e6320 has a 7 mult which is pathetic for overclocking, and the 8 mult isn't bad, but the e4400 will be the same price pretty much, totally making it not worth it for overclocking. And the 4mb cache? Bah, the cache gives it a 3% gain over the regular e6300 and you won't be able to oc as high as the e4400, so I can't justify buying that, but that's me.


Hope this helps 8)


Only one problem.
Allendales don't clock as high as Conroes. There is a point in buying a Conroe class chip, too bad they cost more.
April 11, 2007 2:45:58 AM

Quote:
But if you get better cooling in exchange for the worse ocing for the same price, then you'll still get as high, but with safer temps, and also with the e4400 and watercooling, I think 4ghz is possible


Good point. Pity I don't do watercooling... yet. I'll have to look watercooling sometime.
(Voted E6420: pain to hit over 450FSB, and 8x mult = 3.6 GHz @ 450, which is my target clock)
April 11, 2007 3:01:50 AM

Quote:
I think you can hit 3.6ghz on an allendale, but that's me, what do I know, I'm a taco :lol:  But cheap watercoolers are improving and the prices are dropping, check out the thermaltake big water 745 for one, it's taco approved for a cheap alternative to high end watercoolers


I see Allendales topping out at 3.4 at HardForum (http://www.hardforum.com/showthread.php?t=1138241) even on water. :-(
April 11, 2007 3:06:14 AM

Quote:
That's weird, because someone here hit 3.8ghz on air on an allendale :?

Link please? :-D
HardForum has Prime/Orthos stability requirements (8 hours, I think), hence the overclocks over on the big list are fully stable overclocks.
April 11, 2007 3:19:05 AM

Quote:
That's weird, because someone here hit 3.8ghz on air on an allendale :?

3.8GHz on air stable? I'm not going to say impossible, but its pushing it. Thats more than a 100% OC on air alone... :?
April 11, 2007 3:19:13 AM

Quote:
Since it's late, I'm going to bed, but I'll post it here if I can find it again


OK, thanks. 3.8 Allendale? Sweet jesus, forget the E6420, I'm a poor college student.
April 11, 2007 6:19:43 AM

I've seen freaks of nature in the most impossible theory,so yes 3.8 on air does'nt seem that far fetched,especially on a perfectly matched system...It's like the holy grail,but it happens.
April 11, 2007 7:19:04 AM

Quote:
I've seen freaks of nature in the most impossible theory,so yes 3.8 on air does'nt seem that far fetched,especially on a perfectly matched system...It's like the holy grail,but it happens.
True, but it's like saying that human-beings can run @ 22-23mph. Only a handful of 100 yard sprinters can hit that speed, whereas the rest of us couch-potatoes/average-Joes can probably do 15-17mph. :wink: :D 
April 11, 2007 12:19:25 PM

Okay - follow up question...

I have a p5w dh dlx motherboard, and as I understand it, this mobo has some known characteristics as fsb is scaled up:
- hole/wall at approx 370 MHz. This is a chipset limit which can be jumped over, but then you are forced to use spd timings on memory.
- upper fsb limit of approx 450; as this limit is approached, odd memory settings which aren't available in bios must be relaxed and mch voltage must be upped to the point where better-than-stock-heatsink cooling is required

Given this info:
- the memory bandwidth graph in the link I posted earlier
(http://www.thetechrepository.com/showthread.php?t=30)
- I haven't done any overclocking before
- I am on a budget, so wanting to avoid after market cooling on cpu & nb

...I'm thinking it's better for me to stick to fsb's below the fsb hole, use tight memory timings, and get a cpu with a multiplier of 8 or 9:
fsb = 370 MHz with x8 -> cpu = 2.96 GHz
fsb = 370 MHz with x9 -> cpu = 3.33 GHz
(x10 too high on regular intel hsf cooler? -> cpu = 3.7 GHz )

I could then easily run memory (Corsair 6400c4 rev 1.2 micron d9) at 740 MHz @ 3-3-3

So basically what I'm asking is, is my plan alright? What cpu would you recommend for me?

I'm open to advice from anyone here - really appreciate it!
April 11, 2007 5:14:48 PM

If you're on a budget and wanting to get a good OC with cheap cooling you definately want e4300 or e4400. With 10x multi on e4400 you could just stop below 333 mhz FSB to keep your NB strap tight, use DDR2-667 with tight timings for extra performance and still hit 3.32ghz and smoke an e6420 (8x multi, double the cache) @ 370mhz FSB (2.96ghz) at most tasks.

The reason there are FSB "holes" is because the NB stappings are automatic on these mobos (On my Abit IC7G MaxII Advance I can pick any strapping I want from 400, 533, 667 to 800. don't know why these new mobos don't offer a feature I have on a mobo that is ~5 years old). The reason it works at 401mhz but not 371mhz is because the timings are relaxed at 400mhz if the mobo supports that strapping (ie 1600) and impliments it automatically at that speed (some mobos don't support the strapping or impliment them at different speeds which means that the "hole" becomes a "wall" or a "quagmire of sucky performance" if they impliment the 1600 NB strap at 360mhz instead of 400mhz to give you the warm tinglies about getting a high FSB more easily). There is also a strapping at 333mhz (ie 1333) so if you're on a budget and want good performance you should take some benchies at 332 mhz FSB with the tightest stable memory timings you can get, they might be better than what you get at 333-370 (or close enough to make you want to run cooler).

Actual performance > brag-able CPU Speed > latencies > brag-able FSB speed* for most applications

*for most applications there isn't a FSB or Memory bandwidth bottleneck past ~266mhz and you'll get more performance out of tighter latencies than merely increasing bandwidth further

What I'm going to try and do is hit 345mhz FSB on an e4300 by stopping at 332mhz in BIOS and then clock gen up another 13mhz to avoid NB strap penalties for round 3.1ghz (easy to brag about that way xD) with tight NB and RAM timings that should give me a tasty performance edge without having to run high frequencies or voltages. I'll see how close I get. In practice I probably won't notice the difference between that and a demure 2.9ghz unless I run a benchmark program. To go from 322 to 345 is only 7% more FSB and RAM (in 1:1) bandwidth and 7% tighter NB and RAM latencies (iff you don't relax the timing to get the extra speed) with only 7% more core CPU speed. That's why it might hurt performance to screw up your latencies just to make a small jump past 332.

I don't think there is much data on this though. Most of the data is on Conroe OCing into the +400mhz.

Count me in for wanting to see links to reports of allendales over 3.4ghz plz. I didn't think they could do it (at least not without phase-change cooling).
April 11, 2007 6:47:13 PM

Are clock generators all their cracked up to be?I beleive that you have "skinned the cat" in a compound fracture kind a way.Please let us know your benchies and real world applications ,as I for one am very intrested in your outcome.Brilliant I say!Brilliant? 8O :D  8O
April 11, 2007 9:49:41 PM

Yeah I thought the 3.7 and even maybe the 3.33 were pushing it :oops: 
That all makes solid sense now - just one further question on the link I posted...

As you say, mobo's like mine (p5w dh dlx) don't allow manual selection of the strap. I interpretted the first graph in the link as saying that Asus had done some fiddling with when the straps changed over to allow high (>400) fsbs on 975 motherboards which don't have 1600 straps. The fiddling shifted the 1067-to 1333 strap changeover to approx 400 fsb (as opposed to the 333 you said)...
"...around 370fsb Intels 975/965 chipset starts to have issues if its left on the 1067 strap..."

This says to me that any fsb I can get below 400 (the changeover) should keep the 1067 strap with its tighter latencies (which gives the good nb performance you were talking about). That is why I was looking at the 370 fsb number...

Is what I've said above correct, or have I mis-interpretted the graph from the link (again!)? Can you confirm 975 motherboards don't have a 1600 strap?

I'm really appreciating your help Flasher702 - it's really helping clarify things!
April 11, 2007 10:31:58 PM

Officially there is no 1600 strap yet as there is no 400mhz (1600 after teh quad-pump) FSB Intel CPU out. 1333mhz will be out shortly. As per my understanding of posts from other users: Some motherboards appear to have a 1600 strap that automatically kicks in at 400mhz FSB (which would be correct for a stock CPU running at 400mhz FSB) and some appear to have a 1600 strap that kicks in sooner than that to help smooth through the 370-399mhz "hole" to make OCing easier. As per my understanding you shifted the strapping the wrong direction. The stealth implementation of a 1600 strap at 400mhz or earlier is the theory that best explains the FSB hole. The 1333 strap should kick in at 333mhz but most of the l33t OCers have compiled their data on low-multi Conroes so most of the information you see is dealing with the difference between 360 and +401 (and they did it at a time when DDR2-800 ram was hella expensive too... I'm glad I'm not them!). I forget where I saw it but someone did a comparison that was something like 360mhz*9 (3.24ghz) vs. 410*8 (3.28ghz) with a Conroe "x" series and the slower setting was faster at SuperPi because, theoretically, the 1600 strap kicked in.

I'm not familiar with the p5w and it might be prudent to consult a resource that is for detailed instructions if you plan on clocking above ~370mhz. Personally I plan on NOT doing that to save myself the hassle and the expense and with a 9x or 10x multi you don't really need to.

All this talk of strappings and behind-the-scenes settings makes me miss Socket A all over again xD I think I'm going to try and find someone to buy this rig off me when I get it build, uber cheap brisbanes haunt my dreams. If you're on a tight budget THAT is where the action really is. You can get an x2 3600+ and mid-range mobo and then replace the CPU and Mobo again later for less than a single C2D with a high-end mobo. There are quite a few people around who have had good luck with moderate OCs on brisbanes. That usually the kind of strategy I use but AMD didn't bother to warn me that they were dropping their CPU prices :/ 
April 11, 2007 11:07:03 PM

Okay - I think the only question for me now is whether the strap change at 400 fsb (ie. differences in performance from your x series example) is change from 1067 to 1333 strap, or 1333 to 1600 strap.

I'm inclined to believe 1067 to 1333, but I guess the only way to find out is with some testing :D  Using superpi, if I find an increase in test time from 330 fsb to 335 fsb, I'll know there is a strap changeover here (strap changeover reduces memory bandwidth at the change frequency by increasing nb latency). Then 370 must be on the 1333 strap and the 1600 strap exists because of the other changeover at 400 fsb. If there is no bump in times, I'll know 370 is on the 1067 strap...

In any case, I won't be going above 370 fsb - I'll probably end up going for the 4300 now:
1. cheapest of all
2. decent overclock
3. multiplier suits at 370 fsb
Only downsides are the loss of cache and no virtualisation (which I'd probably never use anyway)

Thanks for your input everyone, especially Flasher 702. :D 
April 12, 2007 12:25:06 AM

I chose an e6420 because to me, its just a good value. If I can get a e4400 that would be great, but then there would be no point to buy a good overclocking mobo. for example, ds3 is capable of around 500 (give or take a few) mhz fsb. U aren't even going to reach that on an e4400. Plus, I wanted to get an e6300. Now if I wait until the 22nd, im getting an increase in performance (and a pretty good one at that) for the same price. Dont get me wrong, if they don;t have the 6420, im orderring the 4400. But really, if you buy an e4400, theres nothing wrong with going with some cheap biostar mobo that has a 400 fsb limit.
April 12, 2007 1:21:17 AM

How much does overclocking reduce the life expectancy? If at all?
April 13, 2007 1:42:01 AM

Quote:
No way a ds3 is capable of 500fsb, that way high, perhaps a bit over 400, but not 500, that's either rd600 or 680i status, perhaps gigabyte's dq6 could, but not the ds3. That's the bottom of the line mobo for conroe's based off the 965p chipset


are u serious? theres someone with a sig on thg in this thread (forgot who) who has a ds3 running at 540 or 530 fsb. ds3 is the hottest budget oc mobo out right now.
April 13, 2007 1:51:18 AM

Decided to skip the price drop day for my fathers system.

Got an OEM E4300 from eWiz for 133.90, and Sunbeam CR-SW-775 for 19.99.

CPU wise, still saved 46 bucks over a presumed 66 bucks ($179 - $113).

And besides, should be able to OC to at least 2.4 ghz (E6600) which just enhances the savings. :D 

:oops:  <--twiches . o O (still have ta wait till monday, though)
April 13, 2007 2:04:45 AM

450 - 475MHz for an unmoded DS3 (tested it myself)
!