Sandy Bridge gets thoroughly benchmarked

Clears up some of the confusion, wows in the graphics department

THE FIRST IN-DEPTH benchmarks of Sandy Bridge have appeared courtesy of Anandtech and it looks like Intel has done more than just a few tweaks to Clarkdale and has managed to come up with something quite impressive. Not only is the new CPU core a big improvement over Clarkdale, but Intel has really delivered on its promise to boost the graphics performance by two times that of Clarkdale. Some explanations as to how Sandy Bridge will overclock is also provided, but this is unlikely to make everyone happ

We’re going to try to summarize Anandtech’s 13 page preview here, but we’d recommend that you head over and read the full thing to get a better idea of what’s coming. Let’s start with the overclocking controversy. The rumour was that Intel was more or less pulling the plug on overclocking with Sandy Bridge, well, it turns out it’s not quite that bad, but don’t expect things to stay the way they are. First of all it looks like Intel will be pushing its unlocked K models as the option to go for if you’re an “enthusiast”, although expect to pay a price premium for these models compared to their non K counterparts. The K models will allow the multiplier to be adjusted all the way up to 57x, although we very much doubt it’ll be possible to get anywhere near that figure for now.

As far as the other models are concerned, well, Anand starts off by saying “So long as Intel doesn’t artificially limit turbo modes, we now have the ability to run CPUs at whatever clock speed they can run at without exceeding thermal or current limits.” Considering that the dual core models don’t have Turbo this seems like a very odd statement and Turbo is limited when it comes to using all of the cores. However, this isn’t the fully story, as apparently Intel is thinking about allowing users to manipulate the multiplier on at least some of the standard Sandy Bridge processors to allow for some overclocking, but it’s not clear as to by how much.

With regards to manipulating the BCLK, well Anand puts it as “There will be some wiggle room as far as I can tell, but it’s not going to be much. Overclocking, as we know it, is dead.” So there you have it, no more pushing of the system bus to try to make your CPU faster, something we’re fairly certain is going to upset a lot of users, not just Intel’s idea of “enthusiasts”. Not everyone can afford to buy the high-end special overclocking CPUs – although even in this case you won’t be able to push the BCLK – and as much as Intel doesn’t seem to approve of overclocking any more, this is likely to be a great opportunity for AMD if it can offer processors that are competitive in comparison to Sandy Bridge.

Oh, hang on, we almost forgot, there’s more. There will be three chipsets for Sandy Bridge, P67, H67 and H61. The latter two support the integrated graphics (more on this later), while the P67 doesn’t allow you to take advantage of the IGP. This in itself doesn’t make a lot of sense, but it gets worse. The P67 is the only chipset that allows you to tweak the memory speed independently from the BCLK and as nice as it is of Intel to allow the memory to be pushed all the way up to 2133MHz, this isn’t likely to impress the memory manufacturers as it now means that there’s a more or less artificial cap on how fast memory you can use. The P67 is also the only chipset two support a pair of PCI Express slots for graphics, albeit still limited to two x8 rather than x16.

The P67 and H67 will both support RAID as well as 14 USB 2.0 ports, six SATA ports of which two will be SATA 6Gbps and up to eight PCI Express lanes for additional slots and onboard features such as USB 3.0 host controllers. The H61 is likely to only be popular by OEMs as it only features 10 USB 2.0 ports, four SATA ports and six additional PCI Express lanes. The good news is that the additional PCI Express lanes will all be full PCI Express 2.0 spec, unlike the current generation of LGA-1156 chipset which operates at half speed.

Intel promised – and delivered – up to twice the graphics performance of its previous IGP, at least judging by the benchmarks run by Anandtech. It’s worth remembering that this is with early drivers and reasonably early silicon. It’s not clear of the IGP was using six or 12 execution units, but Anand made the assumption that the CPU he used featured six. Intel has yet to reveal which CPUs will have what configuration when it comes to execution units.

It’s also unclear if Turbo boost was implemented for the graphics in the CPU used for the test. The IGP on all standard models has a base clock of 850MHz, but will turbo to 1100MHz, with the exception of the Core i7 2600 models which will go all the way to 1350MHz. We’re not quite sure why anyone with one of the high-end CPUs would even bother with the IGP, but this might have been a request from some of Intel’s partners.

The graphics performance is by no means overwhelming, but it’s nonetheless impressive, as Intel’s new IGP is on par with AMD’s ATI Radeon HD5450. Anand even goes as far as to say that this might spell the end for entry level graphics cards, but then again, only Nvidia is going to lose any sleep over this, as AMD should hopefully be offering similar – or better – levels of performance from its new on die IGP.

General CPU performance is very good as well with the 3.1GHz model tested beating the current Core i7 880 in just about every single test. Single threaded performance in Cinebench R10 is the best of any of the CPUs on test, including Intel’s $999 Core i7 980X. This might be seen as a poor comparison, but it shows the raw potential of the Sandy Bridge core over Intel’s previous generations of CPU cores. For those looking for better gaming performance, well there isn’t much in it, at least not in the tests run, but Sandy Bridge proves to be just as capable here as in all the other benchmarks.

Power consumption at idle hasn’t really improved, so if you’re looking for something more power efficient you’ll have to go for the S or T models. However, Intel has managed to improve power consumption during load by about nine percent, not a huge improvement, but still an improvement. This also bodes well for the mobile parts which will hopefully show a similar reduction in power usage during load.

So, to sum things up, overclocking is changing or rather going back to the way it once was, something that’s unlikely to please the masses unless Intel manages to offer enough multiplier options for the “hobbyists” to boost their CPUs within safe limits. Intel’s artificial chipset limitations is also another peculiarity, as there’s really no need for both the P67 and H67, it just appears that Intel has done this to be able to offer two different SKUs. The good news is that performance is up, in regards to both the CPU cores and especially the IGP. There are still a few unknown factors, but at least this gives us a good first impression of Intel’s mainstream platform for 2011.S|A

The following two tabs change content below.