I’d just read through the December edition of PC Format (yes, yes I know, its going over to the Dark Side, but its good to see what other people are saying) and I came across something which lead me to some beard-scratching and Wiki-ing. And then counting. And then double-checking, more counting, extra Googling, looking at whitepapers and press releases and then more coffee. Just what was it that had me searching until two in the morning, you ask?
PC Format contributor and TechRadar’s Jeremy Laird wrote in two separate articles published in both PC Format’s December issue and on TechRadar, that you could drop a Trinity-based APU into a AM3+ board and the chip will work normally, just with the graphics die completely devoid of electricity whatsoever – as in, it chops the TDP from 100W to something more palatable at 50-65W for the CPU core. Even PC Format SA’s hardware writer, Michael Reed, said the same thing in another review in the same edition a few pages later. Now, at 2AM in the morning overloaded with caffeine, this screwed with my mind in more ways than you could possibly imagine given that I know a lot about Trinity and socket FM2.
I didn’t believe a word of it initially. I checked out high-res images of the FM2 and AM3+ sockets just to be double-sure (because you never really know, I may be the one in the wrong) and they are pin-incompatible, there’s no way you would be able to plant an FM2 chip into the AM3 or AM3+ sockets unless you cut off or intentionally bend exactly thirty-four pins (voiding the warranty in the process, let me remind you) – its just physically impossible. In my mind, I thought that if Laird was suggesting that AMD FM2 owners try force their chips into a socket that wasn’t physically designed for them, then any hope I had for PC Format boosting its credibility locally, in my view, is lost. That would be a textbook case of bad journalism right there. I triple-check everything I write and I had to read through what Laird wrote four times to make sure I wasn’t seeing things. Thankfully, I’m not on drugs. Neither is he, presumably, because he’s also a motoring journalist and a good one at that.
In addition, this makes me assume that not only do the editors of PC Format not fact-check every one of their articles submitted by contributors properly (and they do fact-check things, don’t get me wrong), but that they also didn’t stop to ask Louis Villazon if something looked a little off because socket compatibility between FM2 and AM3+ was never touted by AMD. The design philosophies of the APU and the company’s desktop chips designed for use with discrete graphics are just too far apart, they even target different markets and price points! A one-size-fits-all socket is planned for the future, but the current two are AMD’s only option at the moment. Even if you managed to match pins to test this theory, the board, I’m betting, probably wouldn’t even get to the POST stage without frying things up a little. Tom’s Hardware famously tried to snap off two pins on an AM2+ chip to fit it into the AM3 socket. It fitted in well enough, but the board wouldn’t boot, while the chip still worked in its original socket. Clearly, though the interface was more or less the same, interoperability wasn’t guaranteed.
Not even socket FM1 boards will accommodate an FM2 processor even though they’re one pin apart, but I’m guessing this has a lot to do with power delivery enhancements as well as BIOS support, because AMD is re-using the A55 and A75 chipsets in FM2 boards that were used for Llano-compatible products. Back in 2011 when socket AM3+ and the new FX chips were introduced, mention was made that the chips would work in older AM3 boards with DDR3 DIMMs, but those boards would require generally better power delivery through extra phases and a reworked BIOS to add in support. Some lucky board owners did get a BIOS update with Bulldozer support (there are even AM3 890GX boards with a BIOS supporting Piledriver!), but it wasn’t going to be the best pairing right from the start.
So anyway, I contacted Jeremy myself through Twitter and asked him about the reviews he wrote this in. AMD doesn’t release whitepapers or technical details regarding its sockets and their respective changes and differences (probably to prevent people reverse-engineering a similar product) and that makes it very difficult for people like myself to check out these facts before embargoes are lifted, much less a reviewer who’s never seen the product before. And below is the short conversation I had with him on Twitter:
Clearly, the reviews I was reading were written just days after the official release of the products previously kept under embargo, so he had had little exposure to the FM2 products beforehand. Jeremy identified the articles with my help and changed them to reflect the correct status of FM2’s un-inter-operability with other AMD-designed sockets. He was quick to reply and check his facts and admitted fault when he saw what the problem was. It looks like it was just a genuine mistake and he seems like a nice guy. I do read his stuff from time to time and it is good.
However, there had to be something that told him that it wasn’t right. After all, he had the physical products in his possession for benchmarking and testing for his reviews and he was/is presumably very familiar with the Llano FM1-compatible products. He’s been doing this review business for a lot longer than I have (I don’t even do reviews that often) and given that no-one knows exactly what pins do what job, merely assuming that it would work is ridiculous. He had to have looked at the pin layout, counted them and realised that not only was it short of AM3’s 938 pin-out, but it also had pins in the wrong places that would prevent it from fitting into the socket. Who doesn’t turn over an AMD chip to make sure the pins are all straight?
Given AMD’s history of sticking to pin layouts for backwards-compatibility only in CPUs of the same family or similar design, he should have known that what he wrote would be an almost impossible task for AMD’s team to engineer, as it would cancel out all their hard work the moment someone bought an APU and married it to an AM3+ board. They already make an Athlon series based on Llano and Trinity products with the GPU die lasered off from the rest of the core logic, so this would also not make sense. And Jeremy’s a smart man, he’s usually a hawk with details. Was this slip a glitch in the Matrix, or did he just not look deeper into it for himself?
I respect him for owning up to the mistake and changing it promptly, but there’s still that nagging question at the back of my head as to where he got that idea from, and how the hell this slipped past the PC Format editors and fact-checkers when both of Laird’s reviews were two months old. Hmmm, this is weird. Maybe I should chalk it up to the silly season setting in. Guess this goes to show its still better to double-check whatever you read even if its coming from an experienced source. Except for NAG, obviously. ‘Cause we rock.
Note: This isn’t an opinion column designed to slag off PC Format or its readers. Now that I’ve gotten a few reader responses and even a reply from Jeremy himself, I realise that this could all be taken the wrong way (and indeed it was, by quite a few people). I’m not in this to dirty the name of PCF, its writers or Jeremy Laird, this is just me trying to figure out why they all believed the same error. My sincere apologies, then, if I stepped on anyone’s toes. Being a dick when you don’t have to be, as I was above, is the wrong move to make for someone that half-way lives in the public eye.
Discuss this in the forums: Linky
I can’t waffle on Twitter. Follow me to see me try.