What a techie 2009 has in store for us

WE HAD fun this year with major product roll-outs, from Intel's Core i7 and AMD Phenom II on the CPU front, to multiple generations of Nvidia and ATI GPUs.  DDR3 memory matured, new netbook platforms were enabled by Atoms and such, not to forget the final Flash SSD arrival in a big way.

Then, late  in the year, the well-engineered "world financial cum economic crisis" hit. The usual way for a select few behind the curtain to again make horrendous profits at the expense of the millions of individual and corporate sheeple who believed in their financial 'structures'. This started eating away at PC market confidence too.

Whether you ask the big US chipmakers, the Taiwanese OEMs, or any channels worldwide, no one dares predict how the 2009 will go in terms of market and sales - all we can do is see what we can expect, wish and ask for the coming year, geeknology-wise.

CPU dogfight - finally on more equal terms?
Assuming AMD's Phenom II roll-out goes well into 2009 with AM3 sockets and higher speed bins towards mid-year, the firm should at least be able to engage in a decent price/performance battle with the remaining Core 2 and mainstream Core i7 CPUs from Intel. Unless AMD can go beyond 3.5GHz quickly,  it won't have that equalisation chance in the high-end enthusiast PC segment - Intel wins there. On the other hand, the abovementioned 'crisis' is expected to hit that high-priced market the hardest in 2009. We'd like to see a much more affordable 3.6++ GHz Westmere Core i7 before December next year.

The workstation and server battlefield will offer more of a rollercoaster ride through the new year: for any memory-bound stuff, the first quarter performance lead will be more on the AMD Shanghai side, as 'Gainestown' Xeon 5500 Nehalem DP platform might not be out till the end of the quarter. Once it does come out though, Intel's performance lead in DP market is expected to be unassailable until far later in the year. Even then, after AMD upgrades its Opterons to DDR3, full 4-link HT3 and such, Intel could simply offer a DP version of its 8-core Becktons to neutralise the performance threat. A no-holds-barred battle between Gainestown-based Skulltrail 2 and a DP Opteron-based Phenom II DDR3 dual socket AMD mobo would heat up the cold boring CeBIT 2009.

Talking about Beckton, we have to wait for it to see Intel regain a lead in MP scalability and  performance. Until then - most probably early fourth quarter of 2009 - AMD scalability, with those HT3 improvements and similar socket structure across both DP and MP platforms, wins there from techie point of view. Once the eight cores of each Beckton, and (hopefully working by then) four QPI links kick in, it'll be interesting to see how the two compare at year end.

Memories - DDR3 prices and voltages down
With all new 2009 PC platforms optimised for DDR3, the time will be ripe for the memory vendors to make that move and have this memory platform finally priced for the mainstream. Not at the same per-MB cost as DDR2 surely, but not more than 50 per cent above either, before the year end. Expect DDR3-1600 to be that mainstream memory - at stock voltage.

The power saving 'green' push will also  see 1.35 v LV DDR3 memory standardisation and first mainboard support earlier than expected. It eases the job for CPU on-die memory controller voltage as well as  allowing the total removal of DIMM heat spreaders.

INQ wish: DDR3-1600 1.35v CL8 memory on all LGA1366/1160 and AM3 platforms before end 2009, cheaply.

SSD: Sleek Slim Darlings or Sweet Smell of Dosh
Crisis or not, having a laptop with a Flash-based SSD or a desktop with a RAID array of those, will still be a sign of prestige - or at least an ability to throw your disks on the floor in a fit of rage and still keep that data - hopefully. While the prices will come down this coming year by at least 3X per gigabyte, these sleek drives will be the darlings of the storage industry for their continued multiple price premium over the HDDs.

We'll see how all those wear-levelling write techniques will help the format to at least outlast the warranty periods for Flash SSDs. No one has used a Flash SSD for more than about a year and a half, and we're curious to see what happens by the end of the year five. If the DDR2 prices continue to plummet, well battery-backed RAM SSDs could be an answer - no worries about write reliability there.

INQ wish: 16GB DDR2 RAM SSD plus 256GB Flash SSD, both sitting on a PCI-E x8 slot, as high-end PC's typical boot drive.

Graphics: Three Cornered Fight Starts - Holding On For a Third Player
No missives from the quaint town of Larrabee these past few weeks, folks - we'd hoped the townfolk would have visited the great metropolis of Singapore for some Siggraph Sling earlier this month, but no one called.

Except for a early workstation unit rollout, we're not sure how much of Intel's upcoming GPU platform we'll see actually shipping in 2009, but it's weight is already being felt in the relevant circles.

And no, it's not the same vapourware weight throwing as the early trailers for that  tearjerking epic Itanic ages ago, whose main historical purpose, looking back now, might accidentally have ended up as murdering Alpha, MIPS and HP-PA and leaving the X86 unchallenged - and Intel does do a tremendous job on the X86 front.

Larrabee, vapour or not this year, is on everyone's radar screen for 2009, especially Nvidia and ATI's. The odd feeling here is that, despite the age-old Intel-AMD match, the latter's GPU group might suffer less from the eventual Larrabee rollout as it has existing friendlier relations with Intel, and at least partial possible product positioning "coexistence" for a while.

INQ wish: a nasty board supporting GeForce GTX300, RV870 & Larrabee on a single mobo at full PCIe bandwidth, with or without Hydra.

On the other hand, Jen Hsun Huang's famed shouting about Larrabee earlier this year may be just a start of a much nastier Intel-Nvidia propaganda battle later next year, as Intel gets ready to move forward.

INQ wish: Nvidia goes not for panic broadcasts, but puts up a product that stands up to Larrabee, even when CUDA goes under OpenCL.

OpenCL and DirectX 11 will be defined as the final 2009 challengers over the GPGPU code, just like the OpenGL does it for graphics vs the Redmond behemoth. Nvidia can and will maintain CUDA for quite a time  - but open stuff, working across multiple chip platforms, is far more likely to take off. Of course, the ultimate approach is GPUs connecting to the CPUs via coherent shared memory models of HyperTransport and QuickPath interconnects rather than PCI, and being seen as 'coprocessors' with inline assembly code if needed, just like the good old X87.

Best wishes, for 2009! µ