I have closed the performance and productivity polls as described in my posts here and here, and the results can be seen in the Polls Archive. As with most of the other polls I’ve run here, the distribution of votes has not changed greatly after the first few days.

It is clear from the very different voting patterns in the two polls that blog nauseam readers are smart enough understand the difference between the two questions. The performance poll has a very clear skew to the “slower” side. This supports the empirical evidence I’ve seen elsewhere that people perceive AutoCAD as getting slower. This is stuff they’ve noticed for themselves, not a few milliseconds here and there.

On the other hand, the productivity poll results show a much more even distribution. The five options are pretty equally represented, except that “a lot more productive” has suffered at the hands of the most popular choice, “a bit more productive”. If you calculate the mean result, it is almost bang in the middle. It’s actually slightly worse than that, but by such a very small margin that it is not statistically significant.

Overall, we can say that the average viewpoint expressed here is that people clearly see AutoCAD as getting slower, but that its productivity has stayed about the same. So, does this let Autodesk off the performance hook? If a slowing AutoCAD is balanced by productivity-enhancing features, does performance matter? In my opinion, the answers to those questions are no and yes respectively.

It’s not a safe assumption that productivity features are balancing speed issues for everybody across the board. A new feature may help some users’ productivity, maybe even a majority of users, but it won’t help everybody. Some new features even harm some people’s productivity, which is one more reason for being grateful that Autodesk generally lets us turn them off (although it has been forgetful about that in some cases in recent years). Performance is one of the many things that impact productivity, but unlike most new features, is something that impacts the productivity of everybody. Even a first-time user has to sit around waiting for AutoCAD to start up, while a fast power user will be rendered less productive, and certainly more frustrated, by relatively small hesitations.

Furthermore, if AutoCAD is about as productive as it was a few releases back, is that good enough? Or should Autodesk be providing noticeably more productivity in return for our Subscription or upgrade payments? If not, why do we continue to hand over our hard-earned dollars? Why do we go through the upgrade process at all, with all its attendant costs, struggles and inconveniences?

Autodesk, please put much more effort into halting and reversing AutoCAD’s performance slide. It doesn’t have to be a competition between performance and productivity. Improve the former and the latter will also improve.

It seems that most of you are convinced that AutoCAD is getting slower, but I’ll leave the poll going for a while longer. But even if AutoCAD is getting slower, does that mean that it’s actually less productive? Do the new features introduced in recent releases allow you to produce more useful work in a given time, despite making you wait from time to time? I’ve added a new poll to see what you think.

I see quite a few comments in various places that say that AutoCAD’s performance has been getting progressively worse by the year. Is this what most people think, or just the viewpoint of a few complainers? Let’s find out, shall we? I’ve added a poll that asks for your opinions. Feel free to comment, too.

Note that this is a poll about raw performance, not productivity. It’s possible (though difficult) to make a program go slower but still allow you to produce more work in a given time, so I’ll cover the productivity angle in a later poll.

This poll is purely about how fast AutoCAD seems to you. How often do you find yourself hesitating, or waiting, or even going for a coffee break, while AutoCAD does its stuff? Is this getting better or worse? If you compare it with an earlier release, does it seem faster or slower? It is to be expected that some things will get faster and some slower, but what’s your overall impression?

If you’ve noticed some normal drafting operations are much slower in AutoCAD 2009 than in earlier releases, try turning off the new Layer Palette and see if the problem goes away. For example, editing viewports with the Layer Palette visible can be completely unworkable. Don’t just auto-hide it, close it altogether.

Another problem presented by the Layer Palette is that any layer changes you make are applied as you make them. This sounds great in theory, but if each operation takes a while to perform then that’s much less efficient than the old method where all changes are made at once when OK or Apply is picked.

I know a non-modal layer interface was a common wish and it sounded like a cool idea, but now Autodesk has actually been kind enough to grant this wish I’m finding I prefer the old method. I generally don’t need access to all that layer functionality all of the time, so it makes sense to only have the interface occupying that big slab of screen real estate when I actually need it. Your requirements may differ, of course.

If you’re a layer Luddite like me, you can use the old interface by issuing the Classiclayer command. Alternatively, if you set the undocumented system variable LAYERDLGMODE to 0, the Layer command will invoke the old interface instead of the new one.

You have undoubtedly noticed the large red A in the top right corner of the AutoCAD window. Personally, I don’t like the look of it. The concept is rather Fisher-Price and the execution is poor. No competent graphic designer would align the top of the red A exactly with the top of its surrounding button area like this:

There are so many examples of poor graphic design in AutoCAD 2009 that the overall visual effect is close to that of a rather amateurish shareware product. That’s not what you might expect of a multi-billion dollar company that can undoubtedly afford to pay talented people to do much better, but it’s a relatively trivial matter. You probably want to know how it works, rather than what it looks like.

What’s living under that big red A? It’s called the Menu Browser, and it’s a mixed bag. There’s some useful new stuff under there, which I’ll cover later. In this post I’ll just describe using it to browse menus. Frankly, it’s not very good at that. The video below shows you why, and here are some notes to go with it.

• On a fast machine, the reaction times are slightly sticky. Users of slower machines will experience some frustration; I have waited over three seconds for a reaction. Just like the Ribbon, it’s the initial click that hurts the most, with subsequent clicks on the red A being rather quicker.
• Because the menu structure is one level deeper than traditional pull-down menus, you need a minimum of one extra click to do anything.
• Because the Menu Browser is artificially limited to a small section of the screen (no, you can’t resize it), the menus don’t all fit. That means you not only need to perform extra actions to select certain commands, you need to perform extra actions to even see them. That is, if you’re browsing the menus, this new interface is markedly inferior to the old one.
• With pull-down menus, you need a click to start with, then you can hover (or optionally click) to burrow down until you get to the command you’re after. In the Menu Browser, you need a click, then a hover (or optional click), then it’s all clicks from then on. That is, it’s close to the exact opposite of what you might expect, it’s not even self-consistent, and there is more clicking required than before.

If you regularly use pull-down menus, you may as well type MENUBAR 1 as soon as you install your new AutoCAD. You will lose a strip of screen real estate, but you will have your old menus back.

If it’s an unhappy experience using the Menu Browser to browse menus, what else can you use it for? Among other things, you can use it to search menus. More on that later.

Testing performance under Vista can be an interesting experience. The trouble is, Vista tries to improve its performance by observing what you do and caching it for later use in case you do it again. This leads to something akin to the observer effect in science, where the very act of observing something has an effect on what it is you are observing.

Every time I test Ribbon tab switching performance in Vista, the results improve. In XP, the worst tab switching time I saw on my Core2 PC was 1.6 seconds for the first exposure of the Tools tab. In Vista, the same thing took 1.2 seconds the first time I tried it, but subsequent attempts (after closing and restarting AutoCAD but not rebooting) gave results in the 0.5 to 0.6 second range.

I tried turning off Vista’s pretty Aero interface and saw improvements of about 0.1 seconds in every measurement. With Aero off and testing a second time, I saw first tab switches from 0.2 to 0.6 seconds and second tab switches between 0.3 and 0.5 seconds. Most switches were around the 0.3 second mark.

While I hesitate to base any firm conclusions on such shaky ground as these rather unscientific tests, I can say this with some confidence: it is possible to have a system running AutoCAD 2009 under Vista where the Ribbon tab switching performance meets in most cases, and is close to meeting in all cases, the 0.3 second mark that most users perceive as instant response. However, your mileage may vary. Correction: your mileage will vary.

One of the things I like least about AutoCAD 2009 (at least in Release Candidate form) is that I find it very “sticky”. That is, I find myself having to wait for an instant here, then again there, yet again over there. Most of my testing has been on a middle-aged Pentium 4 (3.0 GHz dual core - not too ancient), and it is particularly noticeable there. On my newer Core2 machine, things are better.

When AutoCAD 2009 starts shipping, I suspect your perception of it will be strongly influenced by your hardware. Top gun users on slow machines are going to feel frustrated; slower users on fast machines will wonder what the problem is.

I made a video that shows Ribbon tab switching performance. This is an important aspect of the new interface. Because the Ribbon hides tools behind different tabs, quick access to those tools relies on near-instant tab switching. How well does AutoCAD 2009 do at that? Let’s have a look on a fairly quick PC (Core2Duo E6600 with 4 GB RAM, under Windows XP SP2 32 bit). I intend to do the same in Vista later.

Measured on a faster machine and viewed objectively, it looks rather better than my perceptions from the slower machine had led me to believe. The real problem is the first exposure of each tab, because after that the tab contents can be retrieved from cached memory. The Tools tab is tardiest; 1.6 seconds here translates to 3 seconds or more on an older PC. For the most part, the tab switching speed is acceptable after the tab has been exposed for the first time. For most users, a delay of 0.3 seconds between input and response is quick enough to be considered instant, and most tabs switch in a time fairly close to that after the first exposure.

One way of avoiding or reducing tab switching frustration is to make your own custom Ribbon with a home tab full of the things you use most of the time. How easy is that? Unfortunately, it’s not. Using CUI to make Ribbon parts is not a pleasant experience. That interface makes the Ribbon look super-snappy in comparison. The Ribbon’s more complex internal structure combines with CUI’s snail-like performance, bugs, restrictions and design errors both old and new, to make custom Ribbon creation a loathsome task.

For many 3D users of AutoCAD, the ViewCube is likely to be the most useful new thing in AutoCAD 2009. There are a couple of problems with it, at least in the Release Candidate:

1. It does not work in 2D wireframe mode (which is, paradoxically, where 90% of my 3D work is done). You need to choose another visual style before it will appear.
2. It seems to slow things down quite a lot in complex drawings.

That said, if you have more than enough computing power for the drawings you usually deal with, then this is a very, very nice interface addition. You can just pick intuitively on the various parts of the ViewCube and your view of the model changes accordingly. You can also click and drag on the cube, and it will move where you place it, while snapping on to the various 45 and 90 degree view directions when you get near them, in the same way that Polar snap works. There is a right-click menu with a bunch of useful options, including a quick and easy switch between parallel and perspective views.

You can also easily switch between User Coordinate Systems using the little menu thing under the cube. Unlike some of AutoCAD 2009’s new user interface elements, the ViewCube looks like a finished, polished tool and provides a genuine boost to productivity. Beautiful.

Dude, Dashboard’s dead. Defunct. Done. AutoCAD 2009 replaced the Dashboard with the Ribbon. If you type in the DASHBOARD or DASHBOARDCLOSE commands, they are just converted to the RIBBON and RIBBONCLOSE commands, which turn the Ribbon on and off.

If you’re a fan of the Dashboard (and I never was), there is good and bad news. The good news is that you can right-click on various parts of the Ribbon, pick Undock and you get a Dashboard-like floating vertical Ribbon that can be resized and configured very easily in terms of turning panels on and off. You can’t do that with Microsoft’s Office 2007 Ribbon. Performance aside (more on that later), I generally prefer the way the vertical Ribbon works, compared with the Dashboard. But I’m still not a fan.

The bad news is that if you put a lot of effort into customising your Dashboard in 2008, your work is lost. There is currently no automated migration path from your Dashboard to the Ribbon, so it’s time to start again. This involves using CUI, and unfortunately the process is slow and immensely frustrating. This is due to a combination of the long-standing CUI shortcomings and a set of new ones introduced to go with the Ribbon. Buy yourself one of those squeezy stress-ball things, you’re going to need it.

I’ve now tested startup times of various AutoCAD releases under Vista. Here are the results, alongside the XP results for ease of comparison:

Same caveats as before, plus the following:

1. Some AutoCAD releases were not installed on both XP and Vista partitions, hence the gaps in the table.
2. The Vista tests were performed on the same PC as the XP tests.
3. The system had a 1 GB USB key hanging out the back, giving Vista a theoretical startup benefit over XP.
4. It’s very difficult to get meaningful performance results out of Vista because its SmartFetch and ReadyBoost technologies are doing their best to improve performance without user intervention. I may be a geek, but I’m not a good enough geek to be able to tell what Vista was doing behind my back during these tests.
5. Repeated startups of AutoCAD 2009 revealed a gradual improvement in startup performance. Startup times went 17.3, 13.3, 12.5, 12.5, 11.0 seconds.

Make of that what you will. In short, in my tests AutoCAD 2009 startup time in Vista is still roughly double that of its recent predecessors and about ten times longer than older releases like R13 and 2002.

This table shows both the initial and subsequent startup times for various releases. Most of the qualifications and caveats from my AutoCAD 2009 - The Prequel Part 6 - Initial Startup Time post still apply here.

AutoCAD 2009’s subsequent startups are much less slow than its agonising first startup, as to be expected. Windows XP is doing that by caching and reusing recently used parts of memory. Release 12’s old code, running in 16-bit emulation, is not able to take advantage of that. It’s definitely an unfair test of Release 12 on this system.

In my tests, AutoCAD 2009 startups (both initial and subsequent) are about twice as slow as other recent releases. Users of older releases on modern systems can enjoy startup performance over ten times better.

One thing you’ll notice (and dislike) right away with AutoCAD 2009 is that it takes a lot longer to get started. How much longer? About twice as long as recent releases, or about ten times longer than ancient speed demon Release 13. (I bet a 1994 AutoCAD user transported forward in time would be shocked to hear that description being used). Here’s a video that shows what the first startup looks like in a collection of releases from Release 12 to 2009:

Now for the qualifications and caveats:

1. Tests performed on a Core2Duo E6600 PC with 4 GB RAM, under Windows XP SP2 32 bit.
2. This is not a strictly scientific comparison. Only one system restart and settle-down period was performed prior to timing all releases one after the other. Strictly, each test should get its own restart and settle-down period.
3. Anti-virus and other security software was left on, but was not performing a scan, during the tests.
4. This is the first startup time only. Subsequent startup times during the same Windows session are quicker, and this subject may be the subject of further coverage.
5. Vista startup times may be better, particularly if you leave a spare USB key hanging out of your PC and allow Vista to use it to improve system performance. This may also be the subject of further coverage.
6. The AutoCAD 2009 time is for the Release Candidate, not the shipping product. It’s possible that Autodesk managed to wring more performance out of 2009 before the code hit the manufacturing stage.
7. Application startup time should not be used as an indication of overall performance. It’s just one aspect of it.

If you only start AutoCAD once a day, it’s not a significant issue. If, like me, you start AutoCAD dozens of times a day, it becomes rather wearing.