Late to the party: B550...

So by now, you've all read the kerfuffle surrounding the B550 announcements. I was quite annoyed at first but I'm more resigned about the situation now. In this post I'm going to go over why I felt that way and why, although I understand the technical limitations at issue, I do not agree with the way this has all been handled.

Emotions can be a dangerous thing...

Many people were upset over the lack of official compatibility for Ryzen 4000 series on the B450 motherboards as part of the B550 information releases these last few days. I can put myself in this camp too as I had literally just purchased a B450 motherboard and Ryzen 3200G with the intent of upgrading once the R5 4600 (or equivalent) was released. I did this because in the past, AMD used mealy-mouthed marketing phrasing that specified that the AM4 platform would be supported through to 2020.

As a part of that, as far as we know, Ryzen 4000 series will be the last AM4 product - with the only additional specification available to that product line being PCIe Gen 4. This was all known well in advance of this point in time. What was not known was that AMD were specifically referring to socket usage instead of chipset compatibility. So, yes, technically AMD were correct in what they stated but they intentionally and knowingly misled their customer base by using the words they did in order to encourage purchasing of B450 products.

That, in and of itself, wouldn't be a problem - it's quite a usual marketing practice that we've come to accept, if not love. What is the problem is that X570 boards have been on the market since July 2019 and B550 has been delayed at least twice so that it's coming out a year later than it was supposed to. This is the first time in three motherboard generations that this has happened and, as a result, AMD held back critical information about the B550 chipsets capabilities and limitations.

What this meant was that consumers were unable to make well-informed purchasing decisions. B450 was the only "budget" board available for consumers to purchase and use for the last three years and, even up until recently, people didn't know whether PCIe Gen 4 was supported on the chipset or not! As a result, people have been vehemently angry and quite vocal about it.

The 300 and 400 series boards all released within the same time frame. B550 is VERY late to the party...

B550 coming late is inconsequential to how I feel about this, what is the issue is that AMD knew before the release of the X570 boards that the Ryzen 4000 series would not be supported on the existing 400 series chipsets but held off mentioning that for over a year because they didn't have a product to fill the gap: X570 released at very high price points and so was not really suitable for the majority of Ryzen 3000 processor sales.

Gamer's Nexus has a good breakdown of the technical issues that are constraining this support for the 4000 series processors on the B450 boards but there are boards that are capable of providing support and I think that, as a gesture of goodwill, AMD should have provided official support on any board that has a CMOS capable of holding the extra info. This would allow (a key word here) the board manufacturers to utilise the official AGESA and provide official support if they decided to.

However, there are opinions that customers always had the option of buying an X570 board and that they should have just assumed that support was not incoming for the 400 series boards. "A $150 dollar board isn't that expensive". However, as I noted above, the X570 boards are too expensive for the mid tier processors available over the last year. A £160 X570 motherboard (the cheapest I could find) is not appropriate for a £155 CPU - a.k.a. the most popular selling CPU on Amazon!

Even comparison with a B450 board isn't great as many mid/high end B450 boards (with decent support, features and expansion slots) are selling at £100-£120. What person in their right mind would buy an X570 board for a Ryzen 3 3100, 3300X or R5 3600/3600X? Only people who were desperate for PCIe Gen 4 lanes and there's really very little reason to want that capability right now.

Yes, you can just go out and buy a £3000 system. But that is a tiny fraction of the market and that's not a valid argument to make here. "People should just spend more money" is not only socially tone deaf but blind to the realities of the mid-range consumer CPU market.

Really, they should have kept things simple: 300 series supports Zen and Zen+, 400 series supports Zen+ and Zen 2 and 500 series supports Zen 2 and Zen 3. Instead they have this situation (which is good for the consumer, mind you) where the 400 series was overstretched... giving rise to expectations of future compatibility.

Yes, people say that very few people upgrade their CPUs... but at the same time, we've not had this rate of CPU advancement in over 10 years now. Zen 1+ to Zen 2 and then to Zen 3 have had performance increments we just haven't seen for a decade. Combined with the release of the new consoles which are more powerful than the majority of mid-range PCs in use around the world today, we are looking at a lot of gaming rigs that are going to be outclassed within 6 months and being unable to effectively run modern AA and AAA games within a year. A 6 core Ryzen 5 3600 is likely to be a minimum requirement by late 2021 when console development is targeting the equivalent of an R7 3700X.

As I said above, and at the risk of beating my point to death, the problem is that the B550 boards were severely delayed and their limitations/capabilities were not spelt out ahead of time.

In my case, I would have had two options available to me: 

• Buy an R5 3600X on my B450 motherboard, knowing that I would be keeping it for a few years.
• Not buy anything and just keep suffering with incredibly long encoding times and being unable to play games until the Ryzen 4000 release.

I can tell you what I would have done - I would have taken the second option because I've been advising my family members to do the same thing (wait until the end of the year to upgrade). However, they have systems built in 2017 and they only game relatively infrequently on titles that are not that demanding. They have zero reason to upgrade.

I still have the potential that MSI will update the BIOS of my motherboard to support Ryzen 4000 series CPUs... at the very least, I might expect the APUs to be supported (as they're technically Zen 2 and AMD lists them separately on their support bar chart above). I can't seem to be able to determine how large my CMOS is so I don't know if it's technically possible but the BIOS file size has decreased from 17.5 MB to 14.0 MB from the version in July 2019 to the latest update in April 2020 - so there's definitely some free space to play with.

The other option I have (which is what I mentioned in an additional comment last post) is that I could upgrade to a cheap 3700X or 3800X in a year or so's time. The prices of both those CPUs has already dropped by £100 and I expect to be able to snag one for a similar price to the 3600 - as the 2700 is available for a similar price point. I think that, barring any seismic shifts in computer technology, those CPUs should last as long as my previous rig managed before an upgrade is required.

On jumping the shark...

Make what you will of this...

I mentioned last time that I'd successfully overclocked my CPU to get really quite good single threaded performance. That was true. However, it took a couple of days to notice something that was not quite so desirable. Overclocking with Ryzen Master resulted in a nice, stable 4.06 GHz (4.1 static) clock at 1.3V. I had inputted those values in the BIOS and let the motherboard take care of the overclocking for me instead of having to remember to run Ryzen master every time I booted up.

1.3 V is a bit towards the high end that some people recommend for the Zen 2 process and people tend to recommend a maximum of around 1.375V for the 12LP nm process. I was seeing my temperatures reach 74-78 °C and realised that I was maybe pushing the temps a bit high at that voltage and so I went back and started undervolting in Ryzen Master. I managed a couple of stable settings between 1.2-1.3V at the same frequency but only for a relatively short period of time.

What I did notice, though was that the temperatures were quite a bit below what I had previously been observing. It was at this point I noticed that the motherboard was over-volting the CPU to 1.4V, ignoring any and all settings I was applying in the BIOS screen. Some google searching has shown me that motherboard manufacturers all have lots of bugs when it comes to overclocking and excess voltage being applied and/or settings ignored. It appears that they're all universally as terrible as each other in this regard so I can't single out MSI for this but it does beg the question - why even provide tweakable settings in the BIOS if the user cannot rely on them working?!

So, take this as a bit of a public service announcement (just one more added to the pile)! I've since returned all my BIOS settings to their default values and intend to only overclock within Ryzen Master for the time being. I think the 4.1 GHz/1.3 V combinaton seems pretty stable and a good all-core uplift to the standard 3.6-3.7 GHz I was achieving at stock voltages. I was even considering pushing my favoured core up slightly to see if it can go a bit higher as I have some thermal headroom on the chip but it would be purely an academic interest and not a long-term overclocking solution.

I'm aware of people reporting that the voltages reported in Ryzen Master and other programmes might also be under-reporting the voltage applied to the CPU cores but the difference in temperature during a prolonged all-core load is substantial so I have some faith that I'm not blasting the bejeezus out of my CPU. I mostly plan to run the overclock during certain games and not as a standard procedure.

Best value processor?

This whole single core versus multi core thing has really confused the relative performance of the market! *Big caveat here - the 1600 AF is compared with its RRP, not currently available price. I also had to extrapolate the 1T performance of this processor as it wasn't on CPU Monkey...

Since I purchased my PC and the news of the 3300X/3100 and the B550 news, I've been wondering a lot on what the correct course of action was that I should have taken. As part of that analysis I wanted to see what the relative performance per unit currency was and what the relative value would be (comparing relative performance and relative price to my 3200G for those same processors). To do this I used the cinebench R20 single & multicore scores from CPU Monkey (I'm assuming there's a standardised procedure for performing these measurements) and confirmed them with R15 scores from the same site (not shown above).

I would have take the values from a more certain controlled environment, such as a review, but I didn't manage to find all of the required CPUs in a single (or even multiple) reviews, which is a shame because these are all competing chips in terms of price/performance. I didn't include any Intel CPUs because I'm thinking solely about my personal situation.

Good news first - the 3300X wins at relative value compared to my 3200G when going for a balance between 1T/MT performance. However, the 3100 actually is the best value for money from a straight-up score/currency ratio.

The bad news is that the low-end of the market now has the same problem the high-end did when Ryzen (Zen 1) first came out: Do you go for single core performance or do you need lots of threads? Many outlets have chimed in with the common refrain that the 3300X is the better purchase than the 3100... but if you're really strapped for cash then the 3100 is clearly the better purchase. If you're not strapped for cash (or at least not to the point where an extra £20-40 will make a difference) then why aren't you buying a 3600, which is only a little more expensive... and if you're buying a 3300X, for the same price you can get a 2600 (the 1600 AF is currently up to £130-ish).

Yes the 3300X has a better single core performance (actually THE best out of all the processors compared here) but some outlets noticed issues with frame time consistency in games which the extra, slightly slower cores in the 2600 probably won't exhibit. Six cores and 12 threads also mean a better ability to perform multiple tasks at the same time or more highly thread-reliant parallel processes (i.e. streaming, encoding, etc.).

The 3200G, despite being an APU, actually has a decently comparative single threaded performance ratio to all the other CPUs - beating out the 1600 AF and 2600. Obviously, it can't really compete on multicore performance as it's the only part without hyperthreading.

The worse news is that this part of the market is now saturated with products that differ very little from each other to the average consumer and will ultimately mean people might be buying the wrong product for them. The small price differences* mean that these products are stacked too closely together and most of them do not reflect the real price they should be being placed at.

*£77 between the lowest to the highest CPU compared here and only £36 if you remove the R5 3600!

As I said last time, I felt that the 3300X is too close in price to the 3600 and the 3100 is too expensive compared to the 3200G and (in theory) the 1600 AF and the 2600. Whilst virtually every outlet trumpets the awesomeness of both new low-end CPUs (they are undoubtedly awesome!) they do not represent good value for money when you look past pure numbers... for that, I recommend the Ryzen 5 2600 or 3600 - especially if you're on a B350 or thinking of purchasing a B450 board.