What is Digg?36 Comments
- inactive, on 10/12/2007, -0/+14FFS. Quad CPU's? I just got a core 2 duo damnit.
That plus DX10 GPU's. Man. My wallets gonna be cryin in a couple of months... - nofxjunkee, on 10/12/2007, -0/+10if your core 2 duo does everything you need and want, then just use it and be happy you have a kick ass processor.
- geminitojanus, on 10/12/2007, -0/+8"So I really have only one question after reading that... Since all these multi-core cpus seem to be sharing the L2 cache for every 2 cores, would they benefit substantially if they all had their own? If so... should we expect to see quad core cpus, each with their own caches?"
Intel's dual-core implementation features a shared L2 cache for a few reasons: loading a singular L2 cache uses less bus bandwidth (instead of two processors asking for difference sections of memory at a time, it only looks as if one processor is asking the memory controller for RAM, and thusly uses bandwidth serially instead of in parallel like other designs). It also allows the processors to run cooler thanks to the ability to resize the L2 dynamically; one processor is able to take over 100%, 75%, etc. of the L2 cache as needed, instead of needing to have a larger L2 all of its own. Lastly, because of the way that the L2 is laid out in physically smaller cells (8 cells of 256kB yielding 2MB L2, 16-cells of 256kB yielding 4MB in more expensive models), it can turn off cells not in use.
It's actually a significant improvement over previous designs of dual-core processors with their own L2s. Intel's future quad-core processors will contain a unified L2 allowing all 4 Quad cores to access it; AMD's next generation processors are using a unified L3 (which feeds into the L2, which feeds into the L1 which feeds the registers), however, they can afford the added latency due to the integrated memory controller. AMD gives up the granularity control of their L2s and makes up for it in granularity of the processors in general (being able to turn off, or power down an entire CPU core, including its individual L2), at the cost of making that L2 space now unavailable to the rest of the cores which could use the additional space to buffer more RAM from the main memory.
So in fact, you should expect the opposite. The idea of adding cores to the computer is to make use of additional resources and the sharing of those resources. Keeping processors with their own caches gives up a huge part of the control over your resource management. However, it does take a significant amount of work to even allow for a processor to share a cache (as most processors write to a "back-side bus" unit [hence the terminology "front-side bus" referring to the processor's interface with the system memory controller] that controls the cache, and this system wasn't originally designed with very much respect to a multi-processing environment, such as having another processor write to the same bus). - dawgma, on 10/12/2007, -0/+7That's the first time I've ever been able to read a CPU review from beginning to end... granted, I've been paying more attention lately as to how a CPU works and what all these benchmarks are, but I thought this was a great review. It made a lot of sense and sounded quite fair!
So I really have only one question after reading that... Since all these multi-core cpus seem to be sharing the L2 cache for every 2 cores, would they benefit substantially if they all had their own? If so... should we expect to see quad core cpus, each with their own caches? - iZealot, on 10/12/2007, -0/+6Yea be happy. Some of us are running Athlon XPs with out SSE3 making it impossible to run some software like the latest Adobe Premiere. Oh and render times on a 2500+ plan suck when i can use video editing software.
Im a poor college student what am I guna do? Oh yeaaaa starve for a few months then get a MBP. I might be able to splurge from time to time for ramen. WOOT! - inactive, on 10/12/2007, -2/+7No now that i've experience the high of going from a p4 550, to a Core 2 duo, it's like an addiction to upgrade.
Someone get me the hell away from my computer so i can buy food. - bnewbie, on 10/12/2007, -3/+8Not an exhaustive list but here are some other reviews posted.
Anandtech: http://anandtech.com/cpuchipsets/showdoc.aspx?i=2866
HardOCP: http://enthusiast.hardocp.com/article.html?art=MTIxMywxLCxoZW50aHVzaWFzdA==
ExtremeTech: http://www.extremetech.com/article2/0,1697,2049896,00.asp
Tom's Hardware: http://tomshardware.co.uk/2006/11/02/kentsfield_released_uk/
Tech-Report: http://techreport.com/reviews/2006q4/core2-qx6700/index.x?pg=1
No Ars article at the time of posting this though - still, should help you get a balanced overview :) - geminitojanus, on 10/12/2007, -0/+5"most programs only use first two cores. "
Uh, what crack are you smoking? Programs don't need to be programmed with explicit numbers of cores in mind, and thusly aren't. Instead, they're programmed to be threaded as they need to be, and let the Operating System abstract away the number of cores. In this case, most modern OSes will completely utilize all available cores as best as possible, given their schedulers and the tasks running.
Running apache? Watch it quickly spread to all 4 cores. Running a text editor? Probably not going to need very much in the way of threading (I say probably, because there are some nifty text editors that can allow collaborative editing and thusly could make use of threading for the networking component). - dawgma, on 10/12/2007, -0/+4@geminitojanus
Thanks for your reply. My head has just exploded and I need to go pick up the pieces.... - 5Clicks, on 10/12/2007, -1/+5Great review, Carl! This review definately shows the quality and thought processed testing that goes on at hardcoreware.net. Not only did you put the Q series up against the E series at stock, but you bumped the QX6700 Extreme to 2.9ghz, putting a clock to clock comparison between it and the EX6800 Extreme. Great job!
- starblazer, on 10/12/2007, -0/+3For people that just uses word processing, spreadsheets and maybe a game of solitaire... then ya... "most" porgrams for those people don't need more than even a single core. Obviously, this is not targetted to "those" people.
- mitrovarr, on 10/12/2007, -0/+3I doubt it. Most of the multi-threaded CPU benchmarks showed a benefit of exactly 2x from the transition from 2 to 4 cores; that tends to suggest that there wasn't any kind of bottleneck and the shared cache wasn't a problem.
Of course, it's always possible that it's an artifact of the benchmarks, and having a cache for each core would improve things. It's difficult to extrapolate that from the more real-world tests, because they could be bottlenecked by other factors entirely (RAM speed, FSB, incomplete multithreading, you name it.) - geminitojanus, on 10/12/2007, -0/+2Actually it should read "dice", as "die" is the singular form.
- mitrovarr, on 10/12/2007, -1/+3I probably shouldn't mention that there are also dual processor motherboards out for Core 2s, which will no doubt be made compatible with the new quad core processors quite soon. Mmm... 8 core goodness. SLI a couple of DX10 cards on that and you'll be set!
- geminitojanus, on 10/12/2007, -0/+2There's actually a very small inaccuracy embedded into that article (and thusly your comment): Intel's power management scheme actually doesn't allow any one processor to be turned completely off; there is always a portion of the processor running at an incredibly low power-state, which keeps a few other units on inherent to its design; the watchdog-power management system is located behind the first stage of its pipeline (the "fetch" stage), which means that it keeps a bit more of its core running than it probably should [basically, due to the way P6 is designed, it _has_ to be this way, otherwise the processor would never be able to awake from a extremely low sleep state such as C4; Core is designed differently (it's essentially a remodel of the entire front (Core 2+) and back end (Pentium M/Core) of the Pentium 3 processor, leaving all of the middle stuff alone), but still keeps part of the fetch unit online for this same reason]. AMD's power management is actually much different and does allow for an entire core's power to be all-but cut (for K8, the power management facility is located in front of the entire pipeline). However, AMD's power management technology is much, much less mature and so far unable to match Intel's power savings.
Both AMD's more recent processors and Intel's processors dating far back include clock gating, with more recent designs being much more aggressive about where gating is applied (gating increases latency, but drastically decreases power consumption by allowing the turning off of complex systems completely). - mitrovarr, on 10/12/2007, -0/+2It'll be just like most processor improvements. First, it'll be something only needed (or really only beneficial) for high end workstations. Then it'll show up on high end desktops, laptops, and gaming machines. Eventually it'll end up on all but the cheapest computers, and finally it'll be something everyone takes for granted (like having an integrated FPU.)
- PYREX, on 10/12/2007, -0/+1uh, no it's not. tomshardware did it a week ago
http://www.tomshardware.com/2006/11/02/kentsfield_released/ - dawgma, on 10/12/2007, -0/+1Okay... well how about this. Using only the *known* limits of current consumer technologies, what kind of setup would provide the maximum performance for a quad core system?
i.e. take the fastest FSB, most cache, the highest clock speed (~3.8Ghz), the highest level of hyperthreading, etc... put them all together - what would the numbers look like?
Why do I ask? I want to know what kind of processing power is potentially available in a quad core system, and then compare it to what is actually produced. - xeal, on 10/12/2007, -1/+2No, it's not. The review from AnandTech includes them too.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2866 - Unbekannt, on 10/12/2007, -1/+2Here is another review, including detailed power measurements of the isolated CPU: http://www.lostcircuits.com/cpu/intel_core2quad/
As for the power management, Intel has the same capabilities of turning off entire cores as well as within the cores turning on and off individual functional areas. In terms of the bus management, the actual memory traffic to the CPU is relatively low, the CPU doesn't really need that much bandwidth, rather the data are fetched based on pattern analysis with a very high efficiency - Unbekannt, on 10/12/2007, -0/+1You are correct, the CPU is never completely turned off, unless the power is completely turned off. It is the same thing as with DRAMs that go into sleep mode where only the internal clock is running for a selfrefresh. Otherwise, it would have to go through the entire BIST and initialization cycle again to wake up. I still think that in terms of functionality this is at least "colloquially" correct.
- ElGuap0, on 10/12/2007, -3/+4This is the only review that includes results from the Core 2 Quad Q6600 that is coming in January
- dhughes, on 10/12/2007, -0/+2 ...and bam! Your Windows OS license is immediately canceled because you're using more than two CPUs.
- geminitojanus, on 10/12/2007, -0/+1So am I, ;)
- gitonga, on 10/12/2007, -0/+0Does anyone know how do these chips compare to Sony's CELL? I heard somewhere that it has 8 cores?!
- codeyman, on 10/12/2007, -1/+18 cores!!!
Are you using your computer to control a rocket or making a high quality 3d animation?? - Kniggit, on 10/12/2007, -1/+1@ geminitojanus:
I'm an engineer, not an english major. ;) - lowlight, on 10/12/2007, -2/+2Cheers, thanks!
- Kniggit, on 10/12/2007, -1/+1The headline is inaccurate.
It should read: "...take two Conroe die and put them in the same package"
CPU != DIE
DIE != PACKAGE
CPU != PACKAGE - GenkiDenki, on 10/12/2007, -2/+1Hexus did it too! :D
- pageofmadness, on 10/12/2007, -1/+0more reviews here, too: http://computershopper.com/roundups/intel_core_2_extreme_qx6700_quadcore_four_cores_to_the_floor
- inactive, on 10/12/2007, -7/+2whats the point? most programs only use first two cores. there are a small number of programs and games that will actually take advantage of this
Browsing Digg on your phone just got easier with our enhancements to the 
© Digg Inc. 2009
— Content created and posted by Digg users is