Nvidia mounts sleaze attack on Intel

15 Apr 2008 | 16:41 BST

By Charlie Demerjian

NVIDIA HAS PUT OUT quite possibly the perfect model for sleazy, disingenuous presentations. It is simply so bad that the firm won't put its name on it.

The presentation, labeled "Intel Graphics.pdf", is the sneakiest below-the-belt presentation from a company known for ethically borderline messaging on the best of days. When it won't sign its name to it, you know it's bad.

When asked about the presentation, Nvidia responded with the following statement. "We distributed this to our sales force as a sales tool for them to share with customers. It is a collection of press clippings from the industry on the increasing inadequacies of Intel graphics in a world that demands more and more visual processing. It was intended as an internal sales tool, not for public consumption."

Sounds good so far, but there are several inescapable problems with that logic. The first is that the entire 44-page PDF has the word Nvidia in it twice, both from the same Businessweek clipping on two separate pages. None of its products is mentioned, not even once, and the backgrounds are flat black, without the usual Nvidia logos or themed backgrounds.

Why is this problematic? If a presentation is "intended as an internal sales tool, not for public consumption", they are clearly marked. They have always have "company X confidential" stamped all over them in a way you can't miss. If the news is embargoed, it is emblazoned all over the place. The aim is for it not to be mistaken for public information, and the disclaimers do just that, clearly.

Nvidia disclaimer

Lately, Nvidia has taken to watermarking every page given to journalists with an identifiable serial number on it. To say the firm is paranoid is unde rstating things. As usual, it doesn't work, but it does give the interns something to do, and makes us giggle every time we realise that they failed.

What I am trying to say is that any official document put out by NV in the last four or so years has several distinctive features. First is a clear background that a five-year old would identify as coming from Nvidia. Second is that confidential slide decks are clearly marked as such. Third, if they are really sensitive, they are usually clearly and individually watermarked.

So the internal use line doesn't stand up, only partly due to the fact that this presentation is not watermarked or identified as Nvidia's own. Our research suggests the firm wanted it out, but didn't want people to know where it came from - it is that indefensible. Even if the document is for distribution to " customers", that is hardly an internal NV group.

How do we know that Nvidia wanted the document out in the public domain? The INQ has seen emails sending this presentation directly from Nvidia personnel with @nvidia.com email addresses to people outside the company. A bit of back-story here, Nvidia does not sell to the public directly, so it has a looser definition of sales person/sales team than many other companies. It may include people at outside companies, and we agree with this broader definition. In any case, the emails sent out were clearly not to people in this category. There is no doubt that this presentation was sent out to slur Intel, not to sell Nvidia, and it was done intentionally.

If this sounds confusing, let us give you an example of the presentation; the first page in fact. Beside it is a representative picture from a presentation given to us back in the old days before NV unilaterally cut us off.

Normal Nvidia presentation front page

Not normal Nvidia presentation front page

Notice the subtle differences? Can you figure out which one Nvidia wants to be associated with? Look closely and clues will emerge. Every Nvidia presentation we have ever seen was clearly identifiable as Nvidia, same with ATI, Intel and AMD. Of the hundreds of documents seen over the years, there was never one in which the company of origin was in doubt. There has also never been one that should have been marked confidential but was not. Ironically, there are lots marked confidential that should not have been, but that is another story.

So far, we know three things: it was sent out by Nvidia to a much broader audience than it stated; Nvidia didn't want it associated with the company, but it is not considered confidential. Some of this Nvidia admits, other bits they do not.

Why is this document so offensive? Part of it is the tone, they do nothing but attack Intel instead of promoting their own stuff. Again, we have never seen a single presentation that did not promote the creator's own products while trashing the competition. This type of attack is unprecedented and quite a revealing black mark on the reputation of any company attempting it.

Lets go through it
Page 1 you saw above, and is simply the words "Intel Graphics", page 2 is the same with "Intel Graphics for Vista?" Page 3 is the quote from Steve Sinofsky about the Broken OS, with the quote “945G barely works [on Vista]” on top, and Microsoft SVP for Windows just below. In the fine print, there is the same quote, all from the famous 'why we realise Vista sucks' MS email thread.

There is no doubt that 945G is pretty bad on Vista, especially considering that the mails were talking about a pre-release OS and pre-release drivers. The OS still doesn't work right, and the 945G chipset is barely adequate for Aero, but the situation has improved since those mails were written.

Even Intel admits the 945G has massive problems under Aero, take a look here. "A less known fact is that the 945G will also run Vista Aero, but the results can be disappointingly riddled with stuttering object motion and dropped frames. "

That quote is about a year old, and things are a bit better now, but still, they knew how bad it was for a long long time. Thoughtfully, Nvidia does not mention that there have been enough driver releases since then to choke a horse, much less highlight any improvements in the intervening year plus. It is almost like they are purposefully ignoring changes that do not support their thesis.

Pages 4-7 rehash the same email threads, moving between i915 and i945, nothing but attack quotes from MS and press clippings from articles about the quotes. They indisputably prove that a low-end, three+ year old chipset that isn't on sale any more can't run a UI it was never meant to, and a two-and-a-half-year-old chipset has trouble with a year-old OS, even though it was supplanted six+ months before the OS came out.

If NV was trying to prove a point, ithe point escapses us, but please do note that they do not make the same claims for current chipsets, mainly because they can't. When reality doesn't match your premise, take the cowards way out and dredge up the past under very dubious and tangential circumstances. Then pray no one notices.

P8 starts out with some relevant topics, it is a page with the words "Intel Graphics for Games?" on it. Following that page is a picture of the Intel white paper about the GMA3000, aka the GPU core in the G965. You can find the white paper here, and there is nothing other than a few quotes from the paper on the page.

This is followed by a clipping from a Businessweek on August 13, 2007. The article entitled Is Your PC A Graphics Wimp? was written by Stephen H. Wildstrom. It basically says that if you need a graphics card, you need Nvida or ATI, if you want to run games. It goes on to use the water in the Sims 2 as an example.

Page 11 is a picture of this, and the Intel G965, lacking T&L in hardware falls back to an older shader model, and does indeed look like crap. Factually accurate, but sleazy beyond words. Why? A lot of reasons.

Lets first start with why NV picked the BW article. The first reason is that obvious, BW is a large, mainstream and well-respected publication. If you can quote it for a mainstream customers, or to train sales people to talk to those non-technical proles, you could pick far less respected sources.

Secondly, they picked it because it was laughably out of date at the time. That is important, the story was current in November 2006, we wrote it up here after we were briefed on it and shown demos by AMD. During that demo, they showed both ATI and Nvidia integrated chipsets. Why Businessweek picked it up almost a year later is an open question, but it did.

The problem with the story being so laughably out of date is that things may change between the time it was current and the time it was rehashed. Between that time, two critical changes were made, the G965 was supplanted by the slightly less bad G35, and the beta drivers that fixed the problem were released.

Exactly a week after the article's publication, the final driver was released, and it fixed the problems. The beta drivers had been out for nearly two months.

Even if the article was correct at the time of publication - barely - it hasn't been true for nearly three quarters now. Basically, Nvidia took an outdated chipset bashed by an article that was 10 months late, and dredged it up three quarters after it was no longer true. What courage that took.

I may call NV PR sleazy, underhanded, disingenuous, truth bending, unethical opportunists, but I will not call them so stupid that they missed the G35 launch. I will also not call them so stupid that they missed the fact that the beta Intel drivers were about a month before the BW article and months before the presentation in question.

They didn't have a leg to stand on, so they cherry picked something from the past that hoping that no one would notice. Sleazy beyond words. One could say they even had dishonest intent, there is no other explanation other than being so stupid that they drool from the corners of their mouth. If this was the case, they would have drowned them long before they finished typing the presentation in. If anyone has a better explanation, feel free to write me.

Pages 12-16 basically rehash the same claims, adding nothing new or current to the provably out of date and inaccurate presentations. How this got by the merest of critical scrutiny is a mystery, unless it was not meant to.

On page 17 we come to another title page, this one with the words " Can You Play Today’s Most Popular Games on Intel Graphics?" The short answer to that is probably not, nor is any sane person attempting to do so. The sole exception to this is the AMD 780G chipset, and that is just barely adequate. Basically, I agree more or less with NVs assertion. So, why is it so problematic? The next page has the following pie chart.

This looks pretty bad

Wow, that looks rough, right? I mean, a third or so of the games run, another third have problems, and another third flat out don't work. Holy baloney, that is bad. So, what is disingenuous about this? Look at this slide, from a presentation labeled GeForce7000_Press_Presentation_Final.pdf. It is dated " September 2007" on the first slide. Take a close look at it, really close.

What are the odds?

The first thing you will notice is that on the first presentation, the working/problems/non-working ratio was 37%/26%/37%, and that has now changed to 36%/27%/36%. It looks awfully familiar though, and given a handful of driver drops, including a major one in late February, things should have changed quite a bit right?

If you read the fine print on the first slide, it says "Intel G33: Intel 15.8.0.1437 drivers" and the second has "Intel G33: Intel 14.10.1283 drivers", and both list "OS: Microsoft Vista 32-bit" as the malware of choice.

If I didn't know better, I would say that the stats compiled for last November were fairly accurate, and short of a rounding error, were simply presented again in March. If I did not know better, I would assume the same thing as the Sims sleazing, problems were fixed, they knew it, and decided pick a set of games that matched their conclusion rather than the other way around.

Then again, they might have just picked a relatively odd number of current games that happened to break the current drivers to within 1% of the older ones. You be the judge: disingenuous sleaze or honest mistake? You could also wonder why they picked the low end G33 vs the G35 for this graph? Could it be because the G35 would show almost 100% working games?

When asked about it, Nvidia sent the list of games that they currently consider 'Top 30', along with the results on a G33, and the results are pretty grim. That said, do you know anyone who buys Crysis, Bioshock, CoD4 or Supreme Commander to run on a G33?

The list is top selling games, and the #1 title, The Sims 2 is fair game for integrated graphics, as are a few others like Zoo Tycoon 2, but most are clearly inappropriate. The G33 is a derivative of the G945, a chipset with fixed function graphics, it does not support certain functions, and if a game needs that, it simply won't work.

Nvidia, by picking on the G945, is saying that games that were designed not to run on the G33 will not run on the G33. How is this news again? Like their criticism of the i915 and Aero, if something is clearly labeled 'will not work unless XYZ', and you don't have XYZ, should you be surprised that it doesn't work?

If Nvidia sold cars, they would roll out their latest SUV along with a sales sheet that compares it against the off-road capabilities of the top 30 selling sedans. Then they would point out how badly a Honda Accord does in deep mud, and knock a Ford Focus for it's lack of ground clearance. Technically, both criticisms are correct, but no one buys an Accord for mud bogging.

Also, note how the second slide has Nvidia logos, product names and color schemes all over it, like the green wording versus the white on the 'no-name' presentation. Another accident?

If you are not sold that this was all done with ill intent in mind, try this one on, another noxious example of the breed. It uses Second Life to prove a point, that point being that Intel graphics don't run Second Life (SL). To prove it, they use a quote from a SL web page.

Looks bad part II

Now, if you type in the link listed, it is broken, the page doesn't exist, but you can find the listed page with a little digging. If you look at the SL system requirements page, it shows that Intel graphics from the i945 on do run SL. The presentation was made, according to the metadata, on March 19, 2008, and references web pages on Feb 29, March 5 and March 10, 2008. This means it was made no earlier than March 10.

The problem? Intel fixed the second life problems with driver 15.8 in late February. Having had the ground move under my feet during an article, I would be more than willing to give NV the benefit of the doubt here, but the facts show they don't deserve it.

Two slides previous to the one shown, they say they tested a long list of games with the SAME EXACT 15.8 DRIVER. It was current enough for them to test at least 19 games on, but not this one. They did, however, know of it's existence, so you have to wonder about anyone who put it into a presentation as a key point without checking the bare minimum facts.

The SL system requirements page, as of March 28th, had been updated to show that Intel chipsets, again i945 and up, worked. The page that NV was pointing to does indeed say what they claim it does, but the authoritative source from Linden Labs says otherwise.

It would have probably been easier to link the requirements page, but Nvidia might claim that they didn't know about it, which may be true. Unless you look at slide 11 from the aforementioned GeForce7000 deck, it looks like this.

Amazing coincidence part II

Gosh, that page looks like a clone of the Slide 20 from the other deck with everything identifying Nvidia scrubbed out! What a coincidence. Their only, albeit very thin, defence to not knowing the 15.8 drivers fixed SL and that the page was updated, is that they didn't know it existed. Well, I guess you might say they did know it existed, and the slide was indeed changed deliberately.

In fact, the only conclusion you can draw from this is that they were using it to bash Intel when Intel was indeed broken, a fair tactic. When Intel fixed the problem, and SL changed the page to reflect that, NV had to hunt down a dubious page to support a knowingly out-of-date premise.

This is by far the most underhanded, ethically corrupt and desperate PR stunt I have ever seen. It is unconscionable.

A quick check of the Wayback Machine shows that the second NV quote is accurate, and in fact, until the late Feb drivers, SL was broken on Intel chipsets. Nothing wrong there. Linden Labs was contacted prior to the writing of this to verify when the system requirements page was changed, but as of now, there has been no reply.

On top of that, you have the classic green font, NV logos all over the place, and a slide that clearly is from the company on one hand. On the other, you have everything scrubbed out, and the characteristic green font strangely absent. Again, almost like they didn't want people to know who it came from.

The presentation goes on and on repeating the same premise, that gaming on Intel sucks. Well, duh. We have been saying that for a while, see here and here and here and here.

From there, Nvidia makes another good point, that Intel has failed to meet a single graphics promise on the last two generations of chipsets. On that, we totally agree, and this article was actually prompted by the slide deck. Intel has a miserable track record, and if Nvidia had simply said this, it would have been more than sufficient. It is correct, and backed up by history, with the added bonus of being currently accurate.

Slide 35 shows some specs on the upcoming G45/43/41 chipsets, with DX10 support claimed on all of them. The title, in white, is "So G35 Didn't work ... But They're Sure G45 Will!!??" This is a valid point, but the real eye opener is that this is a page from Intel's confidential WW02 roadmap, P15 to be exact. I wonder if NV PR or sales was authorised to have that? This is lawsuit territory, or at the very least, a damn good reason for Intel to cut off NV completely, can you say caught red handed?

They end up the train of thought with slide 36, "Vista Premium Requires DX10 this June," followed by "Can You Trust Intel with DX10?" and a picture of the Broken OS box. The short answer? No. If Intel can come out with a working G45 DX10 on day one, hey, great. Until then, trust but verify.

Slides 37 to 44 set out to base bashing of Larrabee on the trust premise. Larrabee is a completely different team, different architecture, and different driver set as G965 and kin. There is nothing other than the Intel name that connects them.

To bash one based on the other is like saying NV 5800 was a miserable failure with a horribly loud fan, so GT-200 has to be awful. It is base fearmongering, but about par for the course in an NV sales presentation.

In the end, what it comes down to is that Nvidia has based it's messaging to the outside world on barely defensible methods and quite frankly, dishonesty. The firm has a reputation among the press that is far and away the worst of any company out there. They are not hesitant to cut you off from samples or information if any organization does not give them a review they deem good enough.

In the end, they have mastered using fear, toys and threats to keep the press in line, and rarely does anyone call them on their egregious behaviour because they want the next hot card.

The problem is that Nvidia seems to have a problem with the facts. They seem allergic to honesty. This presentation is by no means the first one packed with questionable claims, but it is the first in memory that does not even attempt to toe the line.

It is no wonder they scrubbed all evidence of their name off of it, including the metadata. There are at least two things in here that could get them sued. Libel and violating NDAs is not something to be trifled with. Then again, as slide 38 says...

µ

© 2007 Incisive Media Investments Ltd. 2007