VISITING TAIPEI regularly has its advantages. After listening to my questions over the years, by now the vendors automatically bring out their new stuff without prodding. This was the case with the mainboard update a few days ago, as the hit news this month is the upcoming Intel P55 platform for the LGA1156 socket Core i7 / i5 / i3 processors, which are all of Intel's mainstream Nehalem and Westmere offerings for the next year.
The motherboard vendors, from Asus and Gigabyte to DFI and MSI, are all excited about the new platform, although design-wise it is far simpler than the previous ones. Instead of a dual-chip North and South bridge combo for Core 2, or IOH and ICH in the X58 for Core i7, there's now just a single chipset IC connected via the DMI flavour of PCIe, as most of the North Bridge functionality is now in the CPU, and that will include graphics in the Core i3 or whatever Intel decides to call the GPU-enabled 32nm Westmere dual-core CPU in a few months.
So, no more QPI or FSB overclocks or other fancy CPU to chipset link speedups for the enthusiasts. Memory is back to two channels, and yes you will see some really low-power wonders there like the GEIL 1.3 v DIMMs we reviewed here just the other day. Others, like A-Data, will try to sell very high-end DDR3-2200 capable parts there, although I think that may be better suited for the 3-channel X58 Extreme parts. And, while most of the new CPUs for the P55 platform are multiplier-locked non-Extreme parts, the up to four level Turbo auto boost on multiple cores will give a bit more excitement here. Buying a, say, 2.93GHz CPU that in most cases will work at up to 3.2GHz with all cores or 3.46GHz with one or two, doesn't sound bad, especially since the user won't need to do anything to make it happen.
Unfortunately, behind that veil of excitement, most vendors showed concerns too, differentiation from each other being the obvious one. It's pretty hard to add much more unique functionality, as the LGA1156 CPU with the P55 - and later, P57 - chipset integrate almost everything, even all the I/O interfaces and digital display outputs for the future integrated CPU graphics. The reduced number of PCIe lanes not only doesn't allow 2 x 16 PCIe SLI or Xfire, but also doesn't have enough extra lanes to mount sophisticated multilane storage, SSD or network controllers. All that has to go back to the X58 chipset, since after all, this is a mainstream chipset.
Then, we have the "Braidwood" or, in simple terms, that cute little ONFI Flash DIMM slot, which we mentioned here for the first time nearly two years ago. Well, we actually don't have it yet, as Intel disabled it at the end in the P55, leaving it for the P57 early next year. Reasons? Well, on the fly OS caching in flash isn't exactly done 'in a flash'. Intel managers in charge told me that on-the-fly OS caching, especially continued write transaction reliability, and of course OS stability - for Windows of any kind, that's never a given - have to be absolutely okay first.
But then, can one disable caching and run that little Flash DIMM - a very cheap and very fast solution, since the controller is in the P55 chipset - as a mini-SSD boot drive anyway? After all, EVGA's Shamino, the famous overclocker, was hinting at a 16GB rather than 8GB Flash DIMM in that slot, and even overclocking that! Overclocking or not, 16GB is sufficient to keep the OS and maybe a basic desktop suite in memory for instant boot. And, if the caching functionality is disabled, yet the BIOS is allowed to see the ONFI flash port as a boot drive, there's far less testing and validation to do.
There is good news of course. CPU overclocking on this box, whether with locked or unlocked CPUs, should be just as good as for the equivalent LGA1366 Core i7 CPUs. The reduction of memory access to two channels will only impact synthetic memory bandwidth tests, otherwise the results of most applications benchmarks won't really change much. And, it is expected to be a good "green computing" platform, as you'll see some record low-power consumption scores on this. How about less than 80W actual measured at the plug power use for a quad core box with basic discrete graphics? The problem, again, is that these are all already built-in into the basic feature set for everyone. The Taiwanese will have a hard time extracting a further few per cent performance, whether with 2-ounces of copper like Gigabyte and now Asus, or by using super-duper capacitors, chokes and such to improve the electricals and overclocking potential a bit more.
In summary, yeah yet another round of mainboards with ever-lower margins for their Taiwanese makers, and worries about disposing of the old Core 2 board stocks now that the replacement for those is arriving in the P55. Some vendors did grumble about that. Despite the traditional Taiwanese thriftiness that makes the Scottish look generous by comparison, even they are feeling too squeezed here.
Also, there's no new interface benefit here. Just like with the X58, external PCIe chips will be needed for USB 3.0 or SATA-6, so you may actually see those uber-interfaces on the high end X58 mobos first anyway. There will be more margin to be made there, and the Gulftown 32nm six-core 12MB cache Westmere drop-in upgrade in the LGA1366 is just about six months away.
As for the users, yes, the P55 with the Lynnfield chips will give you a very decent power, performance and price combination - lower power than even E-step Core 2 quads coupled with somewhat better overall performance and a Phenom-busting price point. Their upgrade story may be a bit more tricky, as the Sandy Bridge 32nm follow-ons from Haifa labs will come online fairly quickly, just over a year from now, and will use the LGA1155 socket. Oh boy, that one pin will make a whole lot of difference for your new P55 mainboard shelf life. But it will be good for the grumbling vendors' repeat sales, though. µ