Very disturbing AMD article
I read this article about an interview with one of the marketing guys from AMD and it looks pretty grim for AMD fans in the future as it looks like AMD is planning to make all but the FX line of 64bit CPU's impossible to overclock.
I'm not certain if their plans are to make the non-64bit CPU's clock limited as I didn't read the whole article (I wanted to share this right away) but I can't see how they could.
If AMD does do that I'll be very disapointed in them for turning their backs on the market that pretty much keeps them afloat.
I'm not certain if their plans are to make the non-64bit CPU's clock limited as I didn't read the whole article (I wanted to share this right away) but I can't see how they could.
If AMD does do that I'll be very disapointed in them for turning their backs on the market that pretty much keeps them afloat.
0
Comments
Given this, and I could be wrong, I would expect processor sales to be relatively flat or somewhat depressed over the next couple of quarters - maybe not overall but at least for the enthusiast market. Further contributing to this expectation is the fact that there are no new aps out currently requiring the added horsepower that new cpus are bringing to the marketplace. 64 bit Windows is a long ways off, we can point to HL2 and Doom3 but how much are those going to drive sales in the big picture?
/end opinion piece
I don't think AMD can do that because of their economy and if you really care for AMD like many on this site seems to do then why hate them for locking their cpu:s. I they don't then they might not be able to compete with Intel in the same way as they do now and that would be bad wouldn't it.
Is this grim news for overclockers? Yes. If it's true. Our outcry may change this, so don't sell out to Intel yet. I'm not saying we should call jihad on AMD or anything... that's blasphemy
IF this does come true, I promise there will be some hardware site that makes a petition for it. My bet is on SM.com or [H].
Your first point, about upgrading from slower, is 100% valid. I was making a point that there are fewer and fewer people who have computers now who are NEEDING to upgrade, there's nothing here or coming soon that's going to *require* a step up for anyone who's bought/upgraded to a reasonably current cpu any time in the past ~18-24 months. This is (again, beware of opinion here) going to drive a flat/somewhat depressed cpu market over at least a couple of financial quarters, if not a little longer.
Heck, for most people in the above boat, a video card upgrade is WAY better value than a cpu upgrade.
/end part II opinion piece
It’s time to upgrade the computer. Or is it? Everyday there is a barrage of advertising assaulting our sensibilities with reasons to replace that apparently tired and worn out beige box that sits under the desk. System builders would have you believe this in order to pull hard earned dollars from pocketbooks around the world. The power of a new PC holds promise to unleash a new world of learning and productivity and many believe in the power of the processor. Too many cling to the notion that more is better. They are dazzled by lofty processor speeds and misguided in their decisions.
For most the PC is a mysterious box that brings email, the internet, plays games, ingests and disgorges the occasional CD and allows printing of family photos from a digital camera. It doesn’t do much more than that. It’s not that it can’t do more. It’s just that a lot of users don’t do more with it. How many of us have really tapped the abilities of the office productivity suites from Microsoft or Corel? How many of us have truly explored the world of digital media be it touching up a still picture or editing video?
It’s unbelievable but we use our PCs like the myth about how we use our brains. The often quoted saying is “We use only 10 percent of our brains.” The truth of the matter is that we don’t actually use 10 percent of our brains while 90% lies dormant. We use certain portions of our brains with a given task such as walking or talking or recalling a memory. If we were to focus our minds on the task of being an Olympic Gold Medalist runner it would certainly be unlikely that we could instantly run like one. This is the way of the PC. The “brain” or the processor may be faster but it will not result in an ultimately faster game without the rest of the body to match its power. But let’s face it. The majority buy PCs like people buy cars; they buy, they use, they get rid of it.
The observation made in 1965 by Gordon Moore, co-founder of Intel, that the number of transistors per square inch on integrated circuits had doubled every year since the integrated circuit was invented. Moore predicted that this trend would continue for the foreseeable future. In subsequent years, the pace slowed down a bit, but data density has doubled approximately every 18 months, and this is the current definition of Moore's Law, which Moore himself has blessed. Most experts, including Moore himself, expect Moore's Law to hold for at least another two decades. Moore’s Law is a bit of an urban myth and is best described in terms of doubling computing power be it processor speed, hard drive space, memory, etc.
Enthusiasts will feel like AMD is turning their backs on them.
There is no doubt as enthusiasts are rather “thrifty” folk wanting more for less. Why pay the big bucks when you can overclock a less expensive processor to the same or near same level? This “bonus” is good for the savvy consumer but poor economics for the business. The cost of manufacturing the faster processors is not much different from the lesser not taking R&D into consideration. The larger profit margin is with the faster processors. Therefore by locking processors AMD will force the market into an upgrade. Remember that enthusiasts who overclock only represent a VERY small percentage of the buying public. When was the last time your mother or father overclocked the PC?
The gigahertz wars will always be there as the public, for the moment, if firmly entrenched with the idea that bigger is better. So little the public understand about processors. Speed is no longer the determining factor in computing power. It now will become a battle behind the scenes with such concepts as PCI Express and HyperTransport. Devices, as we have seen already with GPU throughput doubling and tripling, will be able to transfer data in larger and faster quantities. AMD pushes into this technology to surround their processor with devices that benefit processor computing power rather than relying upon the processor to tow the line.
AMD is already testing the waters of locked processors with current models. But rest assured the enthusiast will find a way to unlock them and entrepreneurs will produce devices to make it easier. It has been done before and the cycle will continue.
Nice piece, well written.
It will be.
It shall be done.
/me puts his controversial hat on.
As far as price, AMD is already more than $25-75 lower than Intel for comparable CPU's. Also, adding advertising is hard to do while simultaneously dropping prices. Lastly, I'm sure AMD would love nothing more than to add a "name" OEM retailer. The trick is, how do you pry them away from Intel? Right now Intel has the attitude that if you want to carry a large AMD line, then fine. But we'll gouge you even more for our chips.
But I'm sure there must be a better way to do this. For instance they could make it possible to remove the lock in a way that is non reversible. Like snipping off a pin or something (serving both as an unlock feature and as a visual que that the CPU has in fact been OC'ed) . And they could make it into their specs (in much the same way as the PR rating was implemented) that every BIOS should display a warning during post if the pin is missing (making it harder to sell overclocked machines).
Just some ideas...but I'm sure there must be a way to please every crowd here
Someone ought to write a letter to AMD.
What's your point, Mediaman?!
I read < re-read > parsed through filters to no avail.
Rawrr!
I don't want to do some irreversible mod to my chips to overclock them... then I wouldn't be able to get them replaced under warranty...
By locking, or apparently locking (there have always been ways to override the "locks") the CPUs, only those who have firedns who know that this batch code was understated as to what it was, and works faster, and can figure out how to, will be able to, and those folks are more likely to take a risk on the chance that they might have gotten somewhat of a free lunch. Were AMD to advertise OCability, in the US they would be offering an implicit statement that the CPUS were underclocked and then they would have tons of burnt chips coming in for those that were statistically normal or those that were the oddball ones which were worse than normal-- because of this implicit warranty. So, offically, they cannot support this.
But, and I agree with MediaMan here, wholeheartedly, there are always ways to accomplish this unlocking, and the adventurous will find them and boast about them and the knowledge will spread. To truely lock, the mobo, CPU, chipset, RAM, video cards, and all major components would have to be locked, and statisitically this is impossible as a QC thing unless prices at least TREBLE to allow for 99% QC test before sorting-- not gonna happen, so we live with components that on normal average can be 7% variant, and this is center bell curve range and not second or third stat area of a normal bell curve distribution. There will be some that run 20% faster, some that run 20% slower than normal, this is production reality, simply by many components being less efficient (or many MORE efficient) in any one core of the many "in" (actually, mostly on) a wafer.
THAT is why, electronically speaking, REAL locking will never happen. Batch will vary from batch, and even on each wafer core will vary from core-- up to 20%+-. Those that are caught will be sold in better CPUs, because they ARE better CPU cores. Those that slip through, say 4-5%, WILL be capable of running faster inherently than rated speed. By up to 20%, randomly.
Art and luck, and knowledge of cooling, and knowledge of power manipulation, all play factors in this. Bell curve says that 1\3 to 1\2 of all production will be normal center group, rest will vary higher or lower. Always been so. Probably will be so into infinity, that electronics will vary some and there is no way to in fact remove that.
OC will live on. Period. And be unsupported officially, period. AMD, if anything will claim the OC actual results show very good ability to produce good quality and stability of CPU. And random sample driven core sorting will continue.
John-- Did I say 8K folding points by Christmas??? maybe 8.5K to 8.8K....
I know that AMD has to get their ASP up a bit from where it's at, but locking everything but the $700 chips ain't going to do it. In the overall scheme of things, us overclockers are just small potatoes with a small percentage of total processors sold, both AMD and Intel, but I imagine that with AMD's procs the percentage is much higher. There are presently no tier1 oems selling desktop AMD products but there are a bunch of independent people like us who build systems for sale or for friends which use AMD because of the price/performace ratio. I will even sell an overclocked rig, with the buyer knowing ahead of time that it is overclocked and stability and speeds can't be guarranteed. I'm not saying that I wouldn't buy AMD in the future for machines I build for someone, but face it, it's a lot easier to sell Intel to Grandma who checks her email with it because they see all those horrid Dell ads every day that play the Intel jingle.:rolleyes2 And another thing in Intel's favor is the fact that they also produce some pretty damn good chipsets to support their procs. AMD on the other hand, mainly uses 3rd party companies to produce their chipsets, with them producing chipset only at the start of a new product cycle to get it out the door.
As far as figuring a way around the processor multi lock, well people have been trying to do that with Intel procs since around 1998 and still haven't succeeded. If AMD is using a locking scheme similar to Intel's, then you can kiss unlocking the proc's multiplier goodbye, Ageek. And if AMD migrates all their K8 procs to a 250 fsb like Ed at O/C says it seems likely they wil do soon, then you can pretty much write off big overclocks with anything that has a locked multi. With present technology, I just don't see how the mobo manufacturers are going to get stable operation at 300+ MHz fsb speeds.
This is much the same principle that L2 cache enjoyed when moving from daughtercard on Slot 1/Slot A architecture, to socket-based architecture in late 1999. L2 once ran at half the frequency of the CPU, now on-die cache is 1:1 with the CPU frequency.
The process used was simply moving the cache as close as possible to the chip, and removing any sort of chip in the middle.. In this case the chips that negotiated between cache and core.
The same thing applies to memory.
AMD's taken a step in the right direction by putting the memory controller on the CPU. It's removed one bottleneck from the equation, which is the chip in between the memory and the CPU itself. It moved the memory physically closer to the socket, which is another step in the proper direction.
The same thing that happened with L2 cache is now happening with system memory.
DDR packaging needs to change as well. It needs to move away from TSOP packaging to BGA. As you'll well notice, video card manufacturers have already done this to achieve the near-GHz range on their video memory.. TSOP packaging has been eliminated in favor of BGA. Kingmax tried this, but their poor memory fabrication techniques squelched their execution. It could have been a whole lot better.
BGA memory adhesion allows for tighter timings, higher frequencies, lower interference, and lower electrical leakage. It has to happen. I wouldn't be surprised if DDR2 debuted at 533/667 (266.5/333.66MHz REAL frequencies) with BGA.. But I can't say that for certain.
But suffice it to say, the longer the run the signal must make, and the more chips in between origin and destination, the worse off you'll be. The optimal solution would be to set a gig of memory right on the CPU in the form of massive cache-like architecture, and run it at the CPU speed. But our fabrication techniques are a long way off from that. But in reiteration, eliminating the northbridge's memory controller function, and shortening the distance between CPU and memory is a sure step toward high FSB speeds.