Conroe vs. Allendale - 4 MB vs. 2 MB cache - Superpi comparison

[celerity]

Most people are planing to buy the E6600 because of the extra 4 MB cache (in addition you get an extra multiplier), although many people will still buy E6300/E6400 because of the lower price.

I haven't decided which one I'm going to buy myself, so I took a closer look on which CPU got the best price per performance ratio. I browsed around on the internet and collected the already existing data of E6300, E6400 and E6600. I haven't been able to read and correct the whole post but If you see any mistakes or anything else feel free to correct me .


------------ 4 MB cache ------------
X6800 2.99ghz 1066FSB 4MB L2 11x multi $999
E6700 2.66ghz 1066FSB 4MB L2 10x multi $530
E6600 2.40ghz 1066FSB 4MB L2 9x multi $316

------------ 2 MB cache ------------
E6400 2.13ghz 1066FSB 2MB L2 8x multi $224
E6300 1.86ghz 1066FSB 2MB L2 7x multi $183

E6600 35.11$/multi 25% more expensive than E6400 and 34.62% than E6300
E6400 28.00$/multi 7.69% more expensive than E6300 but 20% cheaper than E6600
E6300 26.00$/multi -, 7.12% cheaper than E6400


////////////////////////////////////
// Superpi 1MB
////////////////////////////////////

Some results from 4 MB and 2 MB cache CPUs. These results are made with DIFFERENT setups, therefore these results
are not 100% correct. This is followed by a theoretical time on a 4 ghz frequency. Frequency/4000mhz * current time.

Conroe (4 MB cache)
---------------------
E6600 @ 4055 mhz = 13.02 s
http://www.xtremesystems.org/forums/...d.php?t=106497
Time @ 4ghz: 13.02 s

E6600 @ 4414 mhz = 11.75 s
http://www.xtremesystems.org/forums/...d.php?t=106873
Time @ 4ghz: 12.97 s

E6700 @ 4501 mhz = 11.36 s
http://www.xtremesystems.org/forums/...d.php?t=103390
Time @ 4ghz: 12,78 s

X6800 @ 5201 mhz = 9.72 s
http://www.xtremesystems.org/forums/...d.php?t=106697
Time @ 4ghz: 12,64 s

Average time at 4 ghz for Conroe: 12,85 s
Average time at 3 ghz for Conroe (approximately, based on the results above): 17,13 s



Allendale (2 MB cache)
------------------------

E6300 @ 2877 mhz = 18.95 s
http://www.xtremesystems.org/forums/...106685&page=14
Time @ 4ghz: 13.64 s

E6300 @ 3381 mhz = 16.59 s
http://www.xtremesystems.org/forums/...106685&page=13
Time @ 4ghz: 14.02 s

E6300 @ 3549 mhz = 16.74 s
http://www.xtremesystems.org/forums/...106685&page=14
Time @ 4ghz: 14.85 s

E6400 @ 3904 mhz = 15.75 s
http://www.xtremesystems.org/forums/...106685&page=14
Time @ 4ghz: 15.37 s

Average time at 4 ghz for Allendale: 14,47 s
Average time at 3 ghz for Allendale (approximately, based on the results above): 19.29 s


////////////////////////////////////
// Superpi 32MB
////////////////////////////////////

I didn't find so many results on 32 MB superpi (feel free to post them if you have any) but so far I use these two.

Conroe (4 MB cache)
---------------------
E6600 @ 3744 mhz = 13m 56.20 s
http://www.xtremesystems.org/forums/...d.php?t=106285

Time @ 4ghz: 13m 02.40 s
Average time at 3 ghz for Conroe (approximately, based on the results above): 17m 23.04 s


Allendale (2 MB cache)
---------------------
E6400 @ 3409 mhz = 20m 21.31 s
http://i57.photobucket.com/albums/g2...87x7-32mx2.jpg

Time @ 4ghz: 17m 21.00 s
Average time at 3 ghz for Allendale (approximately, based on the results above): 23m 07.80 s

Since these results are based on different rigs with different motherboards and memory it's not possible to give an exactly result on how the CPU will perform. Anyway, here's the average result on the tested CPU's.

SuperPi 1MB
===========
Conroe @ 4ghz: 12,85 s
Allendale @ 4ghz: 14,47 s

Conroe @ 3ghz: 17,13 s
Allendale @ 3 ghz: 19.29 s

Difference: Conroe 12.6% faster


SuperPi 32MB
===========
Conroe @ 4ghz: 13m 02.40 s
Allendale @ 4ghz: 17m 21.00 s

Conroe @ 3ghz: 17m 23.04 s
Allendale @ 3 ghz: 23m 07.80 s

Difference: Conroe 33.1% faster


////////////////////////////////////
// Applications and games results
////////////////////////////////////

Here's some results I got from Anandtech.com, Hardware.fr, Bit-tech.net and Xbitlabs, which all compared the two L2 caches.

Anandtech.com
http://anandtech.com/cpuchipsets/sho...spx?i=2795&p=4



Comment: The extra cache mostly effect media apps such as DivX, WME9, iTunes and some games. The average game performance boost is 2,8%, but as you see it varies pretty much depending on what game.
Average results: 3,5% on everything, 2.8% on games.

Hardware.fr
http://www.hardware.fr/articles/623-...o-dossier.html



Comment: 7.2% performance boost for WinRAR.
Average results: 2.9% performance increase overall, 5,5% on games (there was only two games though).

Bit-tech.net
http://www.bit-tech.net/hardware/200...cessors/5.html



Comment: I believe this says it all:

The amount of L2 cache does appear to make a difference - virtually
across the board - but the difference isn't as big as some were making it out to be. In the application benchmarks,
we found that there was a difference of around 2-3% on average. Both the large and small file compression and
encryption tests really benefited from the additional cache, yielding performance improvements of between 6-9%.
In our two multitasking tests, we found that the 4MB L2 cache-equipped Core 2 Duo was around 2.5% faster, depending
on the background task running. With MP3 encoding running in the background, we fond that the 4MB L2 cache Core 2
Duo was nearly 4% faster than it's 2MB L2 equipped sibling.
At low resolution, we saw varied improvements. Call Of Duty 2 yielded virtually no performance improvements at
either 1024x768 or 1600x1200 with high details, while both Quake 4 and Half-Life 2: Episode One displayed larger
performance improvements at 1024x768. However, when resolution was increased, only Half-Life 2: Episode One showed
any kind of performance advantage on the Core 2 Duo with 4MB of L2 cache.
Overall, the Core 2 Duo with 4MB of L2 cache is quicker than the 2MB Core 2 Duos at the same clock speed, but there
are several instances where there is little-to-no difference in performance.

Average Results: 2D Benchmarks 2.7%; Multi-tasking 2.6%; Gaming Benchmarks 1.7%.

Xbitlabs.com
http://www.xbitlabs.com/articles/cpu...o-e6300_2.html



Comments: SuperPi 0.9%?
Average results: 2.2% is average performance boost, 3,9% in games.


////////////////////////////////////
// Conclusion
////////////////////////////////////

I would say Anandtech had a nice summary of what I think describes the advantage/disadvantage with the extra
cache.

The 4MB L2 cache can increase performance by as much as 10% in some situations. Such a performance
improvement is definitely tangible, and as applications grow larger in their working data sets then the advantage ofa larger cache will only become more visible. Unfortunately, you do pay a price premium for this added performance and future proofing as the cheapest 4MB L2 part is the E6600 priced at $316.

This is basically confirmed by the results from different sources.

If you're the type to upgrade often, then the extra cache is not worth it as you're not getting enough of a present day increase in performance to justify the added cost. However, if this processor will be the basis for your system for the next several years, we'd strongly recommend picking a 4MB flavor of Core 2.

The average performance gain on all tests here is 2.2% (this means Allendale is 97.8% of the Conroe's speed) for the 4 MB cache Conroe, in games round 4.1%. You get a few FPS more which means you won't see any difference at all. I don't think it's worth paying $100 more (E6400 vs E6600) for 4 MB cache since the performance difference is so small. I would rather recommend you to save the money for a better video card or something else that would boost the performance more than a few percent.

I'm not excluding that the cache may have some affect on future apps and operating systems such as Vista, but there's still a long time until the releases of these apps. My guess is that apps will be more optimized for dual core CPUs in the future and the extra cache may have some affect then, but not now. If you'r planning to have that CPU for a longer time, I think it's worth paying for extra cache, otherwise, no.

4 MB cache vs. 2 MB cache
=====================
+ Gives a small performance gain in media apps, some games and a few desktop apps such as WinRAR
+ If you are in need of every performance you can get, 4 MB cache is for you
+ A good choice if you are planning to run a server or something else that would require background apps with high memory need

- There's almost no performance gain
- The cheapest Conroe is more $100 expensive than Allendale, which means that if you'r tight on budget you may want to save that money for something else.
- If you plan to change your CPU the coming year you will probably not need that 4 MB cache since for todays apps it only gives a small performance boost. There will be many other options soon when Kentsfield and the new AMD 65 nm CPUs arrives, so there's always a reason to save $100.

Your comments and thoughts?

[sierra_bound]

The only major problem with your analysis, especially with regard to overclocking, is that most of the results you show were probably done with engineering samples. The retail CPU's may be better or worse. We don't know yet. Or, the retail Conroes may clock better than the retail Allendales or vice versa. You can't assume that ES results are an accurate reflection of what you can expect with retail CPU's.The results so far with retail X6800's have been rather disappointing. But so few people have posted results that it's impossible to draw any conclusions.

[ted3]

The gain of 4MB cache is quite big in some benches, almost 10% is A LOT, so i just cant afford to NOT have 4MB For those $100 you also get higher multiplier, no doubt E6600 is worth the extra $$ for me.

[dogsx2]

It's worth a $100 for the multiplier but I think it will end up more the $100 when retail really starts selling.

[Kunaak]

I want Every possible frame I can get when playing games.

people always make these kinda posts... more versus less cache, low latency versus high latency, 1T versus 2T, more ram versus less, XT vs XTX...

people forget, a PC is a whole machine.
neuter enough parts, and dont be suprised if a person with a "slower PC" can play games better then your "faster" PC...

it all adds up.
cas 3-3-3 here, 1T there, 4 meg cache here, XTX there... 3% differences x 4 = 12% rough guess...

but you get the idea.
it may not be the biggest difference one would hope for, but its enough to where some people like me, would pay for it, cause when in the middle of a game, the last thing I want... is that $40 I saved on my Ram or something... making the game just slow enough to be annoying.

but honestly...
when it comes to conroes and allendales... the real selling point should be the multiplier, and how serious you want to be about overclocking that CPU.
if you want a 4 ghz conroe... your better off with the 10 multiplier, then 7 or 8.

interesting post however.
tons of information gatherd.

[dissident]

for anyone who already owns, say, a dual core athlon or opteron, I don't see why you wouldn't purchase the e6600, for the multiplier if not the cache.

If you're going to spend the money to make such an upgrade over a system that was already quite powerful, it doesn't make sense to buy an e6400. Just keep your AMD system.

If you're upgrading from an old system however and don't care about mega overclock and just want a system that is fast at a good price, the e6400 is a good choice, and will still outperform any system to date... it should give you 3 years of fine service at least before it might start to feel old... and it can still overclock well.

[Phosphate]

The jump from 6300 to 6400 seems like money in the bank because you get prportionally more performance for proportionally more money on a 1:1 basis.

When you get to E6600 its harder to say, but I still think its a good deal.

I think Anandtech Summed it up best: it really depends how long you plan on keeping the CPU. So I wonder how much Kentsfield will cost and will people want to move over to that?

Of course we all know you pay proportionally more for bleeding edge relative to the return.

[Donnie27]

It's the 9 Multiplier I wanted just as much as the 4MB L2. Hell, I wish I have cash for an easy 10 Multiplier on the 2.66 E6700.

Another example, after preaching to me about HDDs with 16MB cache and SATA-II being gimmicks, the same person finally saw my rig and accused me of trying to trick him. He swore I was using Raid of some kind. I had to pull the side panel off to show him I had only one drive installed. Now he's using 3 of them.

[redpriest]

Cache is going to matter at higher clock speeds. Let's face it, no one who gives a damn about performance is going to run that conroe at 1.86 or 2.13 ghz. They are going to push it as much as you can. At high clock speeds, the difference between the memory clock and core clock speed is high. That means your memory latency with respect to the core clock speed is going to be hindering you - you are going to be burning 200-300 cycles waiting for memory. More cache helps in those situations.

Games love cache. They need as much of it as they can get.

Any comparison made at 1.86/2.13 ghz is flawed. We are going to be running these chips at 3+ ghz.

EDIT: Fixed quote, lol.

[Mako88]

I think Anandtech Summed it up best: it really depends how long you plan on keeping the CPU. So I wonder how much Kentsfield will cost and will people want to move over to that?

I will be, no reason not too due to Kentsfield being a drop-in replacement on all Conroe-ready motherboards. So for me it's pointless to even think about anything above an E6600, and even that's stretching it...

[[XC] DragonOrta]

OMFG, you quoted his ENTIRE post.

[Turtle 1]

Hi donnie . I really think donnie you should go for the 10X multipier. Ya i know its your money. But hindsight can be an ugly thing sometimes. Also donnie I wish you could wait for more M/B's to be released befor you jump onboard. I know you demand value for your $$ . But sometimes waiting a little while can be very rewarding. Now after saying that . My wife and myself are very happy with the bad ass M/B . But for my Kentsfield I really believe that the RD600 is worth the wait. ( I may even go with the 4core woodcrest just not enough info now to make a sound decision.)

[Mykou]

E6400 @ 3904 mhz = 15.75 s
http://www.xtremesystems.org/forums/...106685&page=14
Time @ 4ghz: 15.37 s

Average time at 4 ghz for Allendale: 14,47 s
Average time at 3 ghz for Allendale (approximately, based on the results above): 19.29 s

MY superpi 1M 6400@3.00 Ghz is 16.884 sec but I have 4Mo L2 and my ram 800@940 ( ratio 4:5 )

[JoeBar]

Why are u guys expecting more for 2mb extra cache?

In my opinion 6600 still remains the best solution for the hw/oc'er enthusiast.

[tys90]

In my position, getting a E6400 instead of E6600 would mean a better video card, which is a better investment IMO. I plan on OCing, so to get the most value I would need a good board....which is the main problem with getting a E6400 right now. So, for people like me with a high-mid end range budget for a whole new computer, I think a E6400 might be the better choice.

[railer]

I want Every possible frame I can get when playing games.

people always make these kinda posts... more versus less cache, low latency it may not be the biggest difference one would hope for, but its enough to where some people like me, would pay for it, cause when in the middle of a game, the last thing I want... is that $40 I saved on my Ram or something... making the game just slow enough to be annoying.

I feal exactly same way

[m0da]

Great research and overall conclusion, thanks!!

[Verisimilitude]

I want Every possible frame I can get when playing games.

it may not be the biggest difference one would hope for, but its enough to where some people like me, would pay for it, cause when in the middle of a game, the last thing I want ...[is for] the game just slow enough to be annoying.

Same here. I want to soak every FPS out of my 2gigs of ram, x850xt and my e6600. If I cap CSS at 100FPS, when I play de_inferno I better get a constant 100FPS. I'm sick of lagging and getting picked T side Banana.

[Phosphate]

I will be, no reason not too due to Kentsfield being a drop-in replacement on all Conroe-ready motherboards. So for me it's pointless to even think about anything above an E6600, and even that's stretching it...

I'm thinking the quad core may be rediculously powerful. Heck we may even need SLI/Crossfire to divide up some of that power

[yotomeczek]

My scores
1M 3568MHz 16s485
32M 3247MHz 16m36s
both single Pi run

[misteroadster]

which m/b yotomeczek ?
I've done same time about with the gigabyte DS3 with 3380Mhz.

[Mykou]

My scores
1M 3568MHz 16s485
32M 3247MHz 16m36s
both single Pi run

nice fsb

but to compare our 6400's it seems ratios and L2 cache does make a huge difference ( @3.1 my superpi 1M is 16.188 and L2 is 4Mo )

Could you post some benches @3.00 ghz to see ?

which m/b yotomeczek ?
I've done same time about with the gigabyte DS3 with 3380Mhz.

what settings, model and ratio did you have for ram ?

[misteroadster]

4/3/3/9 5400 UL 1:1 (483Mhz)

[Mykou]

4/3/3/9 5400 UL 1:1 (483Mhz)

who is the manufacturer of your nice ram ?

[Nephilim]

who is the manufacturer of your nice ram ?

I'd say he has Corsair 5400UL... An old revision with D9DQT (D9 Fatbody) IC's.