IT LOOKS LIKE Nvidia’s mighty mobile business is hitting the skids, and it is going back to competing on price again, begging for deals and renaming the G92 again. The problem is seen with the release of the renamed 3xxM mobile GPUs, and how these are messaged.
Nvidia’s renaming of GPUs is nothing new. The humble G92 chip, a tweaked G80, already two years old, is just the company’s latest move using this business strategy. If you recall, the G92 debuted as the 8800GT, became the 9800GT, and was then renamed to several more 9×00 variants. It was then shrunk to 55nm where it became the 9800GTX+ and a few other lesser chips.
From there, things started to get unethical, and it was renamed the GTS250 to snow consumers. Nvidia’s G200 architecture was a pretty dismal failure, and spawned no variants other than the original shrink from G200 at 65nm to G200b at 55nm. There was nothing below, so Nvidia changed the labels.
A year late, the tweaked versions of the G200 architecture, the G216 and G218 spawned the G210 and GT220 low-end GPUs. These parts were greeted by the market with yawns and ambivalence, and even with their tiny area, Nvidia could not manufacture them. Add in that they were positively trounced by ATI’s prior generation, and things were not looking good. Nvidia didn’t even send the parts to tech websites to review, much less promote the two chips, they were that bad.
The parts did bring marginal video decode, DX10.1, and some power savings to the table, but the mighty Nvidia was crawling in the mud trying to meet the barest minimum spec for Windows 7. The last new chip to be introduced was called the G215, successor to the failed G214. G214 was a G92 that received similar DX10.1 tweaks, G215 was the same thing with only 75 percent of the shader count and size. Nvidia couldn’t make a full G92 update on TSMC’s 40nm process, so it cut the G215 down to get it out the door,
That is where the problems began. The G215 went on the market as the pretty forgettable desktop GT240. It couldn’t even compete with the 2-plus year old G92/GTS250 in performance. Above the G215, there is nothing new, only the now ancient G92 and the EOL’d but not officially EOL’d G200b variants that can’t be made at a profit. Basically, Nvidia had nothing, has nothing, and the successor architecture, GF100 has all the same flaws that sank the G200 variants.
Then came word that the G2xx chips would be renamed G3xx because, well, there were no new chips to legitimately call G3xx. You can see it in the roadmaps from last summer. If your customers are dumb enough to believe it once, why not try it again?
You can almost hear the meetings in Santa Clara, “What do those words the press keeps yapping about, ‘morals’ and ‘ethics’ mean? Time to break out the “3” stickers to place over the “2” stickers, then profit!” The 4 month old G210 was now the G310, and the equally ancient GT220 is now sold as the GT315.
Getting back to the point of all this, we have the mobile line from Nvidia. It recently released a bunch of mobile parts, or at least renamed a bunch of mobile parts. Another spin, from A2 to A3, has gotten bins to the point where the company can make a mobile variant, and that mobile chip is called, wait for it, Ion2!
Yes, the G218A3 is now a chipset because Dear Leader promised a mighty chipset business to the analysts, and Nvidia will have a mighty chipset business even if it doesn’t have chipsets. Genius! Hand out the bonuses! As a hint to Nvidia management, the definitions for those words I used earlier, ethics and morals, can be obtained in almost any dictionary, as can a definition of “chipset”.
That was quickly followed up by a completely new line of G300 based mobile parts. Okay, there are no new parts at all. The top line GTS360M is just a GTS260M with less than 5 per cent higher clocks. The others, the GTS350M, GT335M, GT330M, GT225M, 310M and 305M are similarly scammy, the same chips with new names, and the occasional tweaked spec. When you don’t have anything to offer, spin, and then apply stickers liberally.
This is where the odd parts come in. The new parts are all listed as DX10 even though they are based on chips that are unquestionably DX10.1. Look at the screen shots below. The GTS360M does not support DX10.1 or DX compute according to the product page, but the GTS260M, same chip mind you, does.
GTS360M without DX10.1 and DX Compute
GTS260M with DX10.1 and DX Compute
Now why would Nvidia do this? Could it be that the upcoming ATI mobile parts are going to crush these chips? Most assuredly. What is Nvidia’s response? Well, it has none. The only way it has sales is to crater prices. In early Q4, OEMs told SemiAccurate that Nvidia sales reps slashed prices for mobile parts more than 50 percent, and prices have gone downhill from there.
Since there are no GT300/GF100/Fermi mainstream chips out, or variants thereof, there are most definitely no mobile variants on the horizon. With the chips from that architecture being too hot, not to mention too big and not realistically manufacturable, the outlook for a mobile variant isn’t all that, well, hot, pun intended. Since the G215/GT240/GTS260M/GTS360M is notably down on performance from the older DX10 G92 based GTX260M and GTX280M, Nvidia is in a pickle.
It can meet Windows 7 specs with a DX10.1 part, the G215 based chips, and promptly get pummeled into dust by ATI’s Broadway chips, likely Madison as well, or it can re-introduce the, wait for it again, G92 based GTX260M and GTX280M. Anyone see a GTX360M and GTX380M on the horizon as an attempt to be pummeled slightly less hard? Since pummeling is directly proportional to ASPs in the market, less pummeling equals more money.
That brings up another problem though. The G92 based GTX2x0M parts came out long before the G21x based mobile parts, so their lack of DX10.1 was understandable. Now, Nvidia has, and is promoting DX10.1 as ‘good enough’ even if the world has moved on to DX11. Nvidia has a line of DX10.1 parts with the 3xxM moniker, and no competition for ATI’s mid- and high-end mobile parts.
What do you do? Reintroduce DX10 chips on the top of the line, pull favors with Microsoft to get a sticker that says Win7 compatible, just like the Intel 915 chipset, and hope no one notices. Also you hope no one sues like they did with the 915, but that is another matter entirely. Mushroom the consumers(1) as much as you can.
So, if any of the new 3xx series supports DX10.1 and DX Compute, they all do in silicon, certainly the ones at the top should, right? All of the 40nm products, GTS360M, GTS350M, GT335M, GT330M, GT225M, 310M and 305M do support it unless Nvidia fuses it off or kills it in drivers. The 2xxM counterparts do, and they are the exact same chip, yet none of the 3xxM chips say so on their spec sheets. Can you say marketing sleaze?
Nvidia is lost on the mobile front. It has renamed its product line three times, from 1xxM (January and March 2009) to 2xxM (March and June 2009) to 3xxM (December 2009) in one year. Its best is not good enough to compete with ATI’s middle ground, so it is forced to reintroduce the G92 as a mobile part, and deprecate the specs of all the other “3-series” variants so as not to look blindingly stupid.
The company does not seem to have a clue, parts, or ethics. The consumer is being totally used in a cynical manner here, and Nvidia’s marketing has sunk to a level that the tobacco company reps handing out cigarettes to elementary school children in the 1950s would balk at. What a wonderful company Dear Leader has built.S|A
(1) ‘Mushrooming’ is a marketing term that means feeding them sh*t and keeping them in the dark.
Latest posts by Charlie Demerjian (see all)
- ARM upgrades realtime offerings to v8-R and adds Cortex-R52 - Sep 21, 2016
- Everspin and Globalfoundries team up for embedded ST-MRAM - Sep 15, 2016
- Intel’s Xpoint is pretty much broken - Sep 12, 2016
- ARM adds 2048-bit vectors to v8A with SVE - Sep 7, 2016
- AMD releases Bristol Ridge 7th Generation APU - Sep 5, 2016