How ASUS notebook SNAFU caused a war of words between AMD and nVidia
7/15/2009 by: Theo Valich - Get more from this author
Update July 31, 2009 08:52AM GMT:We have asked AMD's representatives in UK and Canada over the driversituation prior to this story being published. After receiving quick"excellent questions, we'll get back to you ASAP" on July 14th, wehaven't heard from AMD. We have re-sent the e-mail to appropriatepeople on July 28 and we are still waiting for the answer from AMD. Wewill have a new story on Monday.
The original story
There is a war of words going back and forth between AMD and nVidia employees, with both sides nudging media to go one way or another. Even though this article may mean we won't see samples or get invited to press events, we decided to pursue the truth, regardless of how harsh it may sound.
The story goes as follows: on CeBIT 2009, nVidia launched rebranded GeForce 9800GTX chips and named then GTX260M and GTX280M. To us, this is a clear mislead of the market and would be liable for a governmental investigation, but for some odd reason - no competitors opened a shilling campaign against nVidia. The reason is simple - all players are involved in rebranding "old lamps for new ones". nVidia claimed that their rebranded 9800 series cards are faster. ATI released Mobility Radeon 4870, a card that blew previous parts outside the water.
On Computex 2009, Alienware introduced M17x, a notebook combining Intel Core 2 Quad Q9300 processor, GeForce 9400M [e.g. if combined with Atom CPU, call it ION] and two GeForce GTX280M graphics cards. On the other hand, ASUS introduced W90, combining Intel Core 2 Quad Q9000 processor, Intel's P45M chipset and two ATI Mobility Radeon 4870 graphics cards.
In the world of desktops, there is no doubt that Radeon 4870 will eat GeForce 9800 alive, but is that what a notebook customer is going to get? When preparing marketing papers for Back-To-School shopping period, all manufacturers hit e-tailers or retailers and simply purchase competing products, run benchmarks and then issue their documents claiming that their product is the best thing since sliced bread. The usual marketing spiel, folks.
But this time around, nVidia's Brian Burke went on newegg.com and purchased ASUS W90 notebook in its only available configuration, as show on picture below:
ASUS W90 in its only available configuration in e-tail
As you can see, the notebook was equipped with the Intel Core 2 Quad Q9000 processor and two ATI Mobility Radeon 4870 graphics cards [e.g. Mobility Radeon 4870 X2]. nVidia's staff compared ASUS'es W90 notebook against their Alienware M17x by using the same battery of tests, releasing the results in their "Alienware M17x Reviewer's Guide". When the reviewers's guide came out, it sparked an outrage.
ASUS vs Alienware benchmark results
ASUS vs Alienware benchmark results
The reason is quite simple: in nVidia's tests, a system with two GeForce GTX280M graphics cards obliterated ATI Mobility Radeon 4870 X2. The difference ranged from 10.2% in Call of Duty 5 to a massive 276.6% in 3DMark Vantage. As you can see for yourself, the results are quite impossible. Or at least we thought.
Specifications: Did nVidia wrongfully compare the two notebooks?
As you can see with the results, there is a difference between the two notebooks. ASUS W90 comes with a Core 2 Quad Q9000, while nVidia-configured Alienware M17x comes with a $750 upgrade named Intel Core 2 Extreme QX9300. The difference between the two CPUs is not small: Q9000 is clocked at 2.0 GHz and features 6MB of L2 cache, while the QX9300 works at 2.53 GHz clock and is equipped with 12MB of L2 cache.
nVidia's Reviewer's guide claims that they've used the processor that actually doesn't exist, as Intel sells only Core 2 Extreme Quad QX9300. Core 2 Quad Q9300 does not exist, so we feel that nVidia went here and "equalized" the two in order not to show obvious CPU dominance on their side. Brian Burke, PR representative from nVidia came out with the following statement:
"The benchmark results provided by NVIDIA accurately reflect the consumers experience.
The ASUS W90 tested was purchased from Newegg in the only available hardware configuration. The system was received with the operating system already installed and the drivers pre-loaded. A visit to the ASUS website will reveal that the driver tested is the only available driver for the system (see attached image).
The goal was to show "the fastest notebook on the planet", so the best available components where selected for both systems. The Asus W90 is only available in one configuration so we had no other options for a CPU. The fact is that is the fastest CPU option available for that system."
Truth to be told, ASUS W90 comes with more system memory, 6GB DDR3 vs. 4GB DDR3 inside Alienware's system. Since Alienware could have been maxed out with 8GB DDR3-1333 memory, it was obvious that guys "ran out of budget" when it came to comparing the notebooks.
To answer the question: can a 500MHz slower CPU be at fault for this kind of performance difference? The answer is quite simply, no.
The CrossFire question
Major reason for this performance shortfall is the problem with Catalyst drivers, e.g. the actual configuration. According to some media outlets, Crossfire was allegedly disabled, hence you had two GTX280M cards going against a single Radeon 4870. nVidia cried out that they did not touch the driver configuration, and that they are "innocent".
We went and asked questions around, and the answer we got was quite surprising: Crossfire was enabled, but the drivers in question don't come with game profiles that nVidia used.
This is a showstopper, folks. You paid $2300 to buy a notebook with two 4870 cards in Crossfire, and you're sure as heck that you'll play games. By looking at the names of games, you cannot really say that the games were tweaked in nVidia's favor: after all, if you pay $2300 for a notebook, don't tell me you're not interested in running games such as Crysis Warhead, Call of Duty series, or Fallout 3?
Who is at fault here? The answer is not simple, but understandable: ASUS. When we checked for latest drivers, it turned out that AMD's own site will give you ATI Mobility Radeon X1800 as the latest mobile graphics card out there. No Mobility Radeons 2000, 3000 or 4000 series exist. According to amd.com, ATI stopped making graphics cards for notebooks in 2006. We asked questions to Ian McNaughton and Jay Marsden of AMD fame, but received no answer at press time. We reserve the right to update this article with their answer once it becomes available.
The key point here is how is it possible not to have drivers with game profiles inside the gaming notebook? It is obvious to us that ASUS assembled a driver from the AMD OEM site and simply did not include the necessary profiles needed in order to achieve good scaling performance, but AMD is not helping the consumer with their driver site. This is a bad and completely unnecessary situation for AMD.
Whatever happens in the near future, the situation with drivers today is such that a rebranded GeForce 9800GTX will beat the living daylights out of Mobility Radeon 4870, and this is simply not acceptable.
The PhysX question
To us, the only really dodgy issue was the scores achieved by 3DMark Vantage. If you take a look at 3DMark Vantage and Alienware's CPU results, you can see that the result of this benchmark were 29202 for the Intel Core 2 Extreme QX9300 inside Alienware and "only" 7754 for the Intel Core 2 Q9000 inside the ASUS one. In real world, Intel quad-core CPU should run at around 7.55 GHz to achieve that score. Needless to say, won't see Intel's quad-core CPU clocked at 7.55 GHz for a long period of time. The reason for this massive discrepancy lies in the fact that 3DMark Vantage CPU test uses nVidia PhysX physics API.
Last summer, nVidia introduced GPU-accelerated PhysX driver, and there was a huge outrage at the 3DMark scores achieved by users with nVidia hardware. In a statement from Mr. Oliver Baltuch, President of Futuremark Corporation, we learned that "nVidia's PhysX driver had no WHQL certification, thus it cannot have validated status inside 3DMark Hall of Fame". However, nVidia gained WHQL certification and the PhysX-accelerated score is being sanctioned only on official overclocking competitions, where "GPU PhysX" has to be disabled.
But, will consumers disable GPU PhysX? The answer to that is "no". Users that will spend couple of thousand dollars on a notebook aren't "Excel Bob" and "Powerpoint Suzy", but rather power users who want to get what they paid for. GPU PhysX is just another member of the pack, and the fact that competition didn't support developers well enough with their Physics API is their problem alone.
In our opinion, there is nothing wrong with systems using GPUs for PhysX, even inside 3DMark. The reason for that is simple: Summer 2008 saw the arrival of PhysX Wrapper for ATI cards. Even though both companies claimed that they will support Eran from NGOHQ and his team, the truth was that the PhysX wrapper was blocked with the next version of Catalyst drivers [according to Eran]. What makes the matter even more funnier was the fact that Radeon 4870 would eat desktop GeForce GTX280 alive in PhysX test, but let's leave that in "could have, should have" category.
Conclusion
In the end, was nVidia wrong in this case? Looking at all of the facts, no they're not. They are misleading the market with GeForce GTX260/280M, e.g. same names as the GT200-based desktop parts, but in terms of this situation, picking the fastest notebook… reality showed that between Alienware and ASUS, between Mobility Radeon 4870X2 and GeForce GTX280M SLI, winner is Alienware M17x.
Now, if you are asking yourself is that the fastest notebook on the market, the answer would be, quite simply - no. SmoothCreations sells a Core i7/Xeon 5500-based "notebook", but the price is more than double than Alienware's M17x.
Click to expand...
Problems? See this thread at archive.org.
