Sunday, May 14th 2017
On Elmor's Open Letter, or The State of the Industry
A post on Reddit is doing the rounds from user elmor, a well-renowned enthusiast overclocker who works for ASUS' ROG Motherboard R&D - specifically, in the development of overclocking and enthusiast features. In his post, he talks about the posture of some motherboard makers, as well as about the state of the market as is, with some interesting tidbits thrown in.
One of the most interesting tidbits to be gleaned from his post is that from his perspective, overclocking's biggest supporters are Intel and AMD, who "seriously love overclocking and have excellent people pushing it internally"? AMD I understand - two generations now they've graced us with unlocked-multiplier processors. Intel, on the other hand, has locked-in overclocking efforts with their K-series processors, and have recently told enthusiasts that they should stop overclocking their i7 7700K CPUs, so... I'm a bit on the fence with the blue giant on that specific regard, at least when it comes to mainstream overclocking. My locked i5 6400 is doing just great in the overclocking department, mind you - just not thanks to Intel. Interestingly, Elmor also sets NVIDIA "in the corner of shame" because of their "reluctance to help us push the limits of PC hardware and locking things down more and more."
Elmor also specifically calls out MSI - giving the example of their X370 Xpower motherboard - as being a showcase of product development mainly guided by something akin to "slapping LEDs on it and call it gaming." And this is something that can be further developed, so bear with me for a little while.
The number of advertised features on motherboards have been growing exponentially (pardon my mathematically inaccurate statement, but you get the point.) Manufacturers have been more and more tending towards a "checklist" development so as to offer all the same bangs and whistles that their competitors do, while trying to throw in a specific twist of their own. There is an overabundance of features which all beg to be tested, but at the same time, they're mainly the same ideas and features implemented in a slightly different way, worded with a huge amount of marketing sugar sprinkled on top.
It's a race towards the top, with each manufacturer vying for the consumer's attention, which naturally also ends up bringing a feeling of "been there, done that" in regards to motherboards and their features. If one simply counts the number of implementations of a simple M.2 thermal shield, or an on-board Realtek audio chip, or a manufacturer's specific LED implementation and control on a motherboard level... Between May 14th and April 1st, we here on TPU have covered north of 20 new motherboard releases, and I'm pretty sure a few have slipped through the cracks.
This brings about the topic of market - and feature - saturation and overlapping. A quick and dirty check shows there are at least 59 (!) different motherboards which pack Realtek's ALC 1220 audio solution. Each manufacturer, however, has a distinct marketing for their product, be it Audio Boost (MSI), Supreme FX (ASUS), Purity Sound(ASRock), AMP-UP Audio (Gigabyte), Audio Boost 4 (MSI) or some other marketing naming.
This cutthroat competition and rapid pace of product launches, releases and re-releases with added features also ends up impacting review cycles and timing, as you could expect. The fact is that the number of advertised features is just too great to extensively cover, and keeping up with, with the depth we would like. I'd say that TPU's reviews - courtesy of our own excellent cadaveca - tend to go deeper than the norm, but that's also part of the reason why there are relatively few of them.
Baseline quality of any given motherboard, from the most bare-bones model to the highest of the top-end, have improved substantially over the years. This makes attributing review scores - or better, achieving differentiation through review scores - harder. And sure, there is a level of diplomacy involved regarding review scores. Is it the right thing to do to give a 7 to a motherboard because some non-essential features are slightly buggy? Should we award the 7 and "kill" the product's image outright, or be diplomatic - some would say sensible - and attribute a score based on the the delivery and the potential of the product? We've all heard of some bug fixes doing wonders for any given product. And a hypothetical 8.5 with reservations regarding the required fixing of some bugs, or a 7 solely on the basis that the bugs exist, paint completely different pictures. Fairness is a hard descriptor to achieve, but it's what must be sought after.
Imitation is said to be the sincerest form of flattering, and the entire market (not just the motherboard market, mind you), is built on it. Whether or not this is healthy is another matter entirely - companies who invest in new features do so knowing that their competitors will immediately look towards matching and surpassing their own implementation. They may have a head-start, but it won't ever be a significant one - and original design, feature and product development is much more expensive and time-consuming than imitation. Paving the road is the hardest part, not actually riding it.
Source:
Reddit
One of the most interesting tidbits to be gleaned from his post is that from his perspective, overclocking's biggest supporters are Intel and AMD, who "seriously love overclocking and have excellent people pushing it internally"? AMD I understand - two generations now they've graced us with unlocked-multiplier processors. Intel, on the other hand, has locked-in overclocking efforts with their K-series processors, and have recently told enthusiasts that they should stop overclocking their i7 7700K CPUs, so... I'm a bit on the fence with the blue giant on that specific regard, at least when it comes to mainstream overclocking. My locked i5 6400 is doing just great in the overclocking department, mind you - just not thanks to Intel. Interestingly, Elmor also sets NVIDIA "in the corner of shame" because of their "reluctance to help us push the limits of PC hardware and locking things down more and more."
This brings about the topic of market - and feature - saturation and overlapping. A quick and dirty check shows there are at least 59 (!) different motherboards which pack Realtek's ALC 1220 audio solution. Each manufacturer, however, has a distinct marketing for their product, be it Audio Boost (MSI), Supreme FX (ASUS), Purity Sound(ASRock), AMP-UP Audio (Gigabyte), Audio Boost 4 (MSI) or some other marketing naming.
Baseline quality of any given motherboard, from the most bare-bones model to the highest of the top-end, have improved substantially over the years. This makes attributing review scores - or better, achieving differentiation through review scores - harder. And sure, there is a level of diplomacy involved regarding review scores. Is it the right thing to do to give a 7 to a motherboard because some non-essential features are slightly buggy? Should we award the 7 and "kill" the product's image outright, or be diplomatic - some would say sensible - and attribute a score based on the the delivery and the potential of the product? We've all heard of some bug fixes doing wonders for any given product. And a hypothetical 8.5 with reservations regarding the required fixing of some bugs, or a 7 solely on the basis that the bugs exist, paint completely different pictures. Fairness is a hard descriptor to achieve, but it's what must be sought after.
82 Comments on On Elmor's Open Letter, or The State of the Industry
The Titanium VRM is fullfilled with the world famous and non reliable cheapcrap Nikos fets. Now you figure there are plenty review sites that can do entire review by not stating their opinion, they just state half truths and nothing else.
During the last 7 years
How often a motherboard has been reviewed and received an score that was lower than "ok" ?
How many boards had overclocker badge all around it's features but then showed sub-par VRM capabilities ?
How many boards had their power capabilities down rated by the manufacturers and how many of these boards had said deficiency spotted by the review sites ?
How many times the users had detected and suffered the effects of those faults ?
Yeah, if you answer to these questions then everything elmor is talking about starts to make sense.
AMD on their graphics side is a bit worse. They have really strict limits on frequency and TDP ranges when building VGA BIOSes. Additionally the maximum memory frequency is artificially limited by driver & BIOS, I can't understand why.
NVIDIA supports overclocking relatively well by default. The main problems are the limited TDP and voltage ranges available. We're able to build "unlocked" BIOSes for internal use, which is a big help. When it comes to sub-ambient overclocking however, there are a few issue causing us trouble. Requests for solutions which we know are solvable by BIOS has been falling on deaf ears. Additionally as soon as the product is only sold by Nvidia (like Titan X/Xp) there's no support whatsoever. It really annoys me when you know there's so much more performance to be had from a product, but we can't get there because of arbitrary limitations. I guess for most consumers this doesn't really affect you, but this is my perspective.
To the TPU editors, I'm curious to know how you can justify giving a 10.0 score to any X370 motherboard when the platform is in such a broken state? I agree it's really good price performance, but still. www.techpowerup.com/reviews/ASRock/X370_Taichi/15.html
Overclocking on Nvidia has been a crapshoot for me everytime with hitting powerlimits voltage limits and what not non stop and all the other ridiculous arbitrary made up restrictions like how it throttles down to 1.081v when you reach like 60C or something around there and how you can't increase the voltage AT ALL.
Overclocking AMD cards has always been really free for me since yea I guess the set power limit on BIOSes are really hard set but most higher end SKUs have high power limits that you won't ever really hit unless doing extreme OC under LN2 or something. And you can easily just modify the BIOS to increase the powerlimit if you need to and also increase the memory speed limit and even get this...modify memory timings. I mean damn that's way more advanced overclocking than Nvidias approach of dumbing down overclocking and only letting people move some sliders around. Also another thing is watch how most Nvidia reference PCBs are garbage while AMD makes the absolute most overkill PCBs with the insanely powerful VRMs.
Just my past experiences. Not a fanboy or anything. If you guys see how i have dual 480s its only because i use em for crypto mining and the fact that i AVOID hardware that thinks it knows better than me like the plague (read: Nvidia cards).
This seems - off?
If it's perfect as is, then what is the point of doing any updates at all, why not just leave it exactly as is? More over, what does it say about the competing products? The Tai Chi with the issues not only inherent in the platform, but with ASRock's own flavor of dysfunction added can't objectively have a perfect score. And in future, If and when all major issues are fixed from AMD and ASRock side. Surely it would mean the board is better at that point, yet it already had a 10. So exactly what is the point of the scoring, what does it even mean?
This kind of thing is exactly what sends tacit approval to vendors that anything and everything is acceptable. 10/10 or 9.9 or some such scoring that has such a positive bias that is makes IGN game scoring look credible by comparison.
Then again, if board reviews have an overclocking section that simply increases the vcore and cpu multiplier then calls it a day. Then yeah, I guess all boards are near perfect.
I'm interested to read how this is explained as well.You've clearly not overclocked AMD cards on LN2.
constant black screens, driver saving unstable clock settings upon crash (i.e the clocks stick and are loaded on windows start as the driver loads so you can get stuck in a boot loop). Heck, take one of your RX480's and put it under LN2. set 1.3v and see if you can get a signal at all to the monitor?
The mem clock limits baked into those very RX480s isn't easy to remove either.
Not an NVIDIA vs AMD thing, but just making you aware that XOC on AMD has been painful for years on end and it's not gotten better. In fact the Rx480 is the easiest to OC simply because of it's very low clock ceiling.
Be it NVIDIA or AMD, they are not making GPU OC easy at all. The people who would know this the most are the ones involved in XOC more than anyone else, as anything other than that is sliders for both vendors.
Voltage also isn't the issue on air cooling for most GPUs, 1.15v on air for GTX 1080 doesn't add any more frequency than 1.087v. Lots of limitations on NVIDIA GPUs and they've gotten really bad. However these are not going to affect air OC for the most part, but XOC.
Anyway...ive been complaining to asus, msi, giga, nvidia themselves about their limits on cards. This is all nvidia really. These aib's want the cards, you must fit it within thier parameters. What i keep asking aib is why cant yoh say, fine 'no warranty returns' on one level of card like the classy, or matrix, or lightning... something.
But you find few people care about this at tpu... its mkre run of the mill ambient clocking in these parts. ;)
but dont ask me how much mainboards(most through user error=pins espacially socket 1366 or shorting them with a wrong placed spacer) we got back telling us we sold them kaputt.
i just tell this because back then i and a lot other benchers in austria bought there and even with ln2 you dont have to ruin hardware if you dont aim for the last mhz which needs quadruple the volt.
hardcore benchers look after there hardware and know the most time what they are doing because just a small elite is getting sponsored and the rest uses their own money
I'm judging the board makers in board reviews, not the CPU makers in board reviews. I'm available for specific consultation to design staff at any company, under contract. ;)It's not just that. BIOS design is ASUS's forte, and is the one area that few can match. There are also many "fit and finish" design cues that are sorely lacking by other board makers. It's funny because I am working on a review of an ASUS Z270 product right now, and a lot of what ASUS offers specifically in the ROG product lines is ethereal and might not be noticed by people who haven't seen many many boards to notice the differences.I hate to repeat myself, but... there is a difference between overclocking to reach "records", and overclocking in a way that can be used by the average user on a daily basis. A big part of that is BIOS support for things that the end user does not know about or understand, and how difficult it is for the average user to use the products they bought that were designed for "overclocking" by the masses.
That's the thing; the masses cannot OC and make records. That's a niche of a niche. In order to sell products to the masses, they have to meet the needs of the masses, not the niche.
The mem clock limit on RX 480s is 1 BIOS value. It really isn't hard to find and replace once you know where it is. I have 2500MHz BIOS for the RX 480 GTR. The only thing I don't know how to remove is cold slow.
The black screen issue is the display drive circuit not liking cold. Though I think the problems with that at least on RX 480s vary from card to card because my XFX RX 480 GTR didn't seem affected by it even at full pot(not that I could actually get any high scores on the card my mount seemed to fail over and over and over again on that card). The ref RX 480 on the other hand dropped out after a few seconds of Firestrike.
Silicon lottery plus an apparently "artificial" limitation sounds like a challenge to me.
Don't get me started.
AMD limits clockspeeds? Where was that? You can easily mod the BIOS for the memory speeds as well and the signed BIOS is easily patched with a small tool that just requires one click where the green company's signed BIOS can't be worked around with.
Go suck Nvidia's d*ck somewhere else.
This unlike your card wont blow the VRM's up when I use that BIOS, nor does it ship with a warning against using furmark. It also performs better than a pair of 480's in all games at all things, short of heat production. I know this because it replaced a pair of RX480's, which replaced an R9 290, which replaced a pair of 7950's (flashed to R9 280 and heavily overclocked). I also have a stack of Fury's I play with on occasion.
Might want to look into the history of some of the members here before assuming they swing one way or the other. I could care less what color the card is. I have a 4k monitor and want to play games maxed out without sounding like a jet taking off or have every other game not work because it doesn't have a profile.
AMD is no hero here. Both companies limit their products.
My statement still stands and other XOC guys also agree that Nvidia is blocking any kinds of enthusiast overclocking on their cards so what are you trying to prove? You sound like an Nvidia fanboy hence what I said I couldn't care less what your previous cards were.