I've been encoding 10 albums from FLAC with 398a3 now, and they're quite consistently 7-12kbps higher bitrate than 397b2. using V0-new, encoding with foobar. Does anyone have any idea why this is, or am I doing something wrong?

Here is a screenshot. In the status bar on the bottom of each instance of Encspot you can see the average bitrate for the whole album, 210.3 vs 221.7 kbps. The percentage of frames used on each bitrate is also quite different, doesn't this suggest a very different encoder behavior?



Moderation: replaced oversized image with thumb

This is the price to pay for the improved quality.

3.97b2 does have improvements over previous lame versions, but other things became seriously worse. If you look at the problem sample thread there is a category of problem samples which cannnot easily be described as 'rare' with 3.97. Wombat described the problems as sandpaper-like distortions, and it looks like he is able to find these samples on a rather regular basis.

The good thing with the problem sample thread is that the lame devs seriously worked on it, and with 3.98a3 they gave us a solution which greatly improves on these problems. Moreover the by most people highly appreciated lame vbr modes (including aps, ape) had some serious problems with certain problem samples from old versions like 3.90.3 up to 3.97b2 or 3.98a2. This is greatly improved now with 3.98a3, but there is a price to pay.

If you encode with -V0 (a good idea IMO): do you really worry about 10 kbps? Enjoy the improved quality!

I know the lame devs are working on reducing this bitrate 'bloat', but may be you will never see a remarkable reduction at the -V0 level.
To me this is good. Most people think like this: oh, VBR is good cause it uses few bits in easily encodable situations and many bits in situations where encoding is hard. While this sounds good it's wishful thinking. Maybe the encoder can decide correctly in the vast majority of situations whether it is easy to encode or not. But there remains always a certain percentage (maybe permillage) of situations where this decision is wrong. The best to do is provide a safety margin.
In the say 100 to 150 kbps bitrate range people are out for most efficient encodings. That's why the safety margin here should be very small or even non-exisiting. In the 150 to 200 kbps range people are out for very good quality while still expecting to get efficient encodings. Here a safety margin of medium size is welcome. In the bitrate range beyond 200 kbps people are out for best quality, and coding efficieny is of little concern. In this case a rather big safety margin is most adequate. As far as I have understood robert, this is exactly what current lame development is about.

That's why you get higher bitrate. I would welcome it.

I think you've heard of TOS # 8 enough, haven't you, halb27? If you want to make judgements, make them personal. eg : I prefer the older, deprecated versions. Newer users using the search feature might fall on this thread and say "well geez, 3.90.3 is _still_ the best version..." -- while this is simply untrue, and abxing otherwise is rare, and with sensitive users only.

I'm only happy for the quality increase, and I don't mind the size. The reason I wondered was simply because the output seemed quite different with the new version compared to differences between earlier versions.

There are opinions, and there are facts.
It's true I've contributed more than once with opinion (but I have not pretended that they were to be something else). But this here is simply fact. You can read about the sandpaper-like distortions on several threads and how they are improved on with current 3.98a3.

Added:
Different opinions are welcome to me, but what I have come across so often and really dislike are simplificated cites: I have never said: 3.90.3 is better than the current version. And in this thread I didn't even mention 3.90.3. My post was about the very merits of 3.98a3. (Or does this bother you cause you prefer 3.97b2? As for the HA recommendations they are valuable but they are no universal untouchable wisdom and especially a time lag in updating the recommendation can occur. Don't you feel like that as 3.90.3 was recommended so long?)

"other things became seriously worse" is clearly your opinion, not fact.

Ok, 'seriously' is opinion. Should have left it out.

Back on the topic

I used to play with EncSpot as well and I noticed that alpha versions of Lame usually use higher bitrates (I might be wrong though). That should correspond to the "bigger safety margins" mentioned by halb27.
When the version becomes beta or stable, the bitrate distribution looks more "Gausian" like and the bitrate goes down.
So, let's wait for a wile until 3.98b.