AMD 785G Update - Multi-Channel LPCM is not Available

Having to print a retraction or deliver information that is totally opposite of what we presented in an article is something we dread, in fact, it is probably the worst single event a reviewer can imagine. Unfortunately, we have to do that today as our coverage on the 785G chipset contains information that appears to be incorrect now after several conversations with AMD this morning.

Let’s just rip the Band-Aid off quickly here. We have touted, as have many others, that the 785G chipset provides multi-channel LPCM audio output. We went over it in our press briefings from AMD and in previous conversations with them and others. It turns out that in retail boards and with the current 9.8 (8.634a) driver set that an audio block exists which limits LPCM output to two-channels. In other words, the 785G is no different from the 780G/790GX chipsets in this regard.

As of now, our initial preview coverage and review is incorrect regarding this feature set. For most users, this will not matter, as two-channel LPCM is available along with 5.1 S/PDIF audio outputs through the HDMI port. However, for the serious HTPC crowd, the lack of multi-channel LPCM is a huge drawback when selecting a platform that will be the centerpiece of your media center. The Intel G41/G43/G45 and NVIDIA GF8200/9300 series of chipsets fully support multi-channel LPCM output as does AMD’s own HD 4xxx series of video cards.

How did this mistake happen? It is a long story we are still unraveling, but I will provide what we know at this point. One of the first items I checked in testing our early pre-production 785G boards/drivers was multi-channel LPCM output. This feature worked based on the application setup, BIOS release, and driver setup we had at the time. In fact, as part of our testing we captured a couple of screenshots and took a quick photo of our A/V receiver indicating PCM direct output on eight channels.

AMD believes at this point that it was an error on the audio output stream or we were actually receiving S/PDIF audio output but the actual signal (flag) was misinterpreted by our hardware/software setup. Another theory is that we received multi-channel output through the Realtek ALC 889a audio codec on the Gigabyte board, however we did not load the Realtek drivers but it is something we are currently testing.

I firmly believe we actually had multi-channel audio output working correctly, but I have been wrong before and certainly will be wrong again at some point. AMD has informed us that the same two-channel audio block utilized on the 780G/790GX has always been present on the 785G chipset except for their Maui products. We are continuing to work with AMD in regards to our test discrepancies, but for now LPCM output is limited to two-channel operation.

We confirmed it on several different boards today after AMD notified us this morning that we should check this capability again. Unfortunately, after the last driver and BIOS updates on Friday afternoon, I did not complete driver regression testing on the audio side of the chipset over the weekend. A huge mistake as it turns out and one that I sincerely regret and apologize for now.

I had several valid reasons but in the end I cut corners in order to finish the chipset review for today. In hindsight, I should have postponed the article until all of the regression testing was completed. Audio output was one item left off the list; the thinking was if it worked before, it surely would work again. Plus the A/V receiver was generating the desired Direct PCM output flag, but a closer look would have revealed the loss of channels as it turns out. So, the A/V receiver is now out of the cabinet and next to the monitor station.

In case you are wondering about what else was left off the list, I did not verify if core unlocking was still working (it does) or general overclocking attributes of the chipset (no changes). I figured those items would be covered in testing this week for the motherboard article next week, as they tend to be board/BIOS/CPU specific.

This brings another problem that we have in general with rushed releases. Originally, this chipset was going to be released at the end of August and of course, for a variety of reasons, it was pulled up to 8/4 a couple of weeks ago. That still leaves enough time to properly test the chipset, but typical of recent product launches, the drivers were not solid enough for serious testing until last Wednesday and did not fully mature until Friday afternoon.

This also occurred with the BIOS releases. In fact, we stuck to our guns that we would only utilize retail BIOS releases for testing. The three sample boards that AMD shipped for review all contain excellent BIOS’, but they were hand tuned at AMD. Only ASUS has committed that the BIOS release on their review sample boards will be available to the public. The BIOS releases we utilized on the two Gigabyte 785G boards are what shipped on the boards that we purchased from Newegg when they first went on sale.

I say all of this as the late release required the retesting of five AMD boards with 57 different benchmarks, amounting to around 1040 test runs and the additional time required for reviewing all of the video features in depth, several of which were not working right until last week. We had three days to accomplish this task. No excuses for what happened on my part, but at times something has to give in order to meet a major deadline. My calculation on what to skip in this particular instance turned out to be incorrect. All I can say is that it will not occur again.

That said, our conclusion about the 785G chipset remains true, but for now, we cannot recommend it to the serious HTPC owner looking for an all in one solution. Our recommendation for this audience continues to be the GF9300/9400 motherboards paired with an E6300 or E7200 processor.