Intel QX6700 Quad-Core

[D_o_S]

When Intel Core Duo hit the stores, people went crazy. Many wanted to experience the amazing performance for themselves. Now, Intel brings us the QX6700 - a quad-core CPU. Is another revolution about to take place? Or are we going to find just a small increase compared to dual core? What about single threaded applications? Multi-core scaling?

Show full review

 

[Deleted member 3]

Quite a price, I take it that someone who buys it doesn't use a 6800 with 1GB of PC6400 though.

Also Clovertown is commercially available already, when spending 1400 on a CPU paying 200 extra on a motherboard isn't exactly a problem, in that respect I wouldn't say they're a different market. The only plus I see for Kentsfield over the cheaper Clovertown (2,66/1333GHz = $1172) is official support for SLI/Crossfire and perhaps memory performance.

 

[Frogger]

Very nice read liked the test bed not being Top of the line give the 'local joe' the heads up.... again

 

[WarEagleAU]

Awesome review D_o_S. Id love to have reviewed one of these puppies. Not too bad from Intel, taking the second highest cpu and putting two on one die. Nice. Im sure AMD has something a lil better up their sleeves. (AT L|EAST I HOPE SO)_

 

[xman2007]

Quite a price, I take it that someone who buys it doesn't use a 6800 with 1GB of PC6400 though.

exactly my thoughts though this makes me happy as i possibly wont be doing my next build until next year so hopefully prices of processors will have gone down quite a bit and ill have a nice 4x4 or intel e6700 / maybe even a qx6*** variant

good review btw, makes me think looking at the difference with 3dmark 2001 compared to later builds it seems to be more cpu reliant than any other as there was a 5000 point increase with the overclocked quad. i hope amd do have somthing up there sleaves to keep up/ whoop intel once again as they have served me well over the last few years

 

[lemonadesoda]

Nice review D_o_S...but why did you test games at such a low resolution... I can understand that you want to remove the possibility of GPU bottleneck... but in practice, I think it is better to use a resolution that people would actually be using in practice, ie. their 1280x1024 or 1600x1200 TFT.

I'd be grateful for an update to the benchmarks if you still have the system there

 

[lemonadesoda]

IMO cache should be refered to as 4 x 2MB, since NO single core (or thread) has access to more than 2MB.

While the intel solution is clever at scaling, it is incredibly inefficient. If you have an application, e.g. Photoshop, running 4 threads or a filter, then each of the 2MB cache has, essentially, the same data.

Any other cache design would make the cache and memory controller very complex and may add additional latency... and I understand why Intel chose the existing solution... but we should definitely not refer to 8MB cache, but 4 x 2MB.

 

[W1zzard]

IMO cache should be refered to as 4 x 2MB, since NO single core (or thread) has access to more than 2MB.

if i remember correctly at idf an engineer told me that the cache sharing on c2d is more complex than just halving it and giving a portion to each core

 

[xman2007]

IMO cache should be refered to as 4 x 2MB, since NO single core (or thread) has access to more than 2MB.

While the intel solution is clever at scaling, it is incredibly inefficient. If you have an application, e.g. Photoshop, running 4 threads or a filter, then each of the 2MB cache has, essentially, the same data.

Any other cache design would make the cache and memory controller very complex and may add additional latency... and I understand why Intel chose the existing solution... but we should definitely not refer to 8MB cache, but 4 x 2MB.

im not saying your wrong about this but where did you get the information that "NO single core (or thread) has access to more than 2MB cache" ? i just havent seen the same data myself

and what would be the point in putting redundant ram on a chip ? im sure enough techies would know that 2mb is the limit of cache on a chip and surely that would be made common knowledge

 

[W1zzard]

there is your data. rightmark memory analyzer is singlethreaded, yet it can access the whole 4 mb of the cache

 

[W1zzard]

fx-60, 1 MB per core, 2 MB total

 

[lemonadesoda]

Thanks W1zzard. I stand corrected. So its 2 x 4MB then?

 

[W1zzard]

yep

 

[D_o_S]

Nice review D_o_S...but why did you test games at such a low resolution... I can understand that you want to remove the possibility of GPU bottleneck... but in practice, I think it is better to use a resolution that people would actually be using in practice, ie. their 1280x1024 or 1600x1200 TFT.

I'd be grateful for an update to the benchmarks if you still have the system there

Hi,

I'm glad that everyone likes the review so far.

I tried benchmarking games at the resolutions you suggested - 1280x1024 and 1600x1200 - the problem is, as you mentioned, the GPU. It really is a bottleneck, which results in very small differences between CPUs (for example, at 1600x1200, the difference between the E6600 and E6700 is less than 1 FPS), therefore I have chosen to not include these results.

 

[Shredder]

Nice review DOS, I am running the Kentsfield and still in the cooling solution level right now. I added crossfire to the mix , with all the fans rocking and rolling I seem to hover around 50 C, at idle..5 percent cpu and less.
ATI tool to keep the vids at 50 percent fan load and CPU Fan on Performance mode (Zalman)
I looked around alot of places for an operating temperature for this chip, not really excited about OCing it without a thermal cap. Any ideas on the operating temperature or the cap for this one ?