Thread: RIP AMD PC's

Someone just needs to get back to work on increasing Mhz, they have been hovering in the mid 2s to mid 3s forever. Cores can only take you so far, and even then I haven't really seen an increase in cores since I got my i7 back when the 1366 socket came out! Or even Core 2 for that matter.

Size matters. AMD is and was too small to match Intel on R&D and so they have to be more clever and flexible. That means that they have to gamble a bit sometimes. The ATI purchase was a gamble that has been a push. Due to the long pipeline of CPU design, my guess is that when Netburst CPUs started matching and often exceeding K8 CPUs they sent K10 down that road of long pipelines and high clock frequency. Intel on the other hand made their next line of CPU much more like the K8 with its excellent per instructions per clock efficiency. But Intel with its vast resources was able to be much more flexible and react to quicker. So now we have a strange situation where they are both making chips that are more like the competition was making when the rivalry was at its peak. It's like Ford is making Cameros and Chevy is making Mustangs

Sent from the land of bier and bretzels on my Galaxy S3 using Tapatalk

Originally Posted by gehn

Someone just needs to get back to work on increasing Mhz, they have been hovering in the mid 2s to mid 3s forever. Cores can only take you so far, and even then I haven't really seen an increase in cores since I got my i7 back when the 1366 socket came out! Or even Core 2 for that matter.

Clock speed isn't all that important... Jumping into the 4GHz+ range is just too costly in power and heat to be viable for the average market. People are free to overclock their machines (and do) if they really want to pinch every inch of power they can out of their machine.

I used to consider myself a pretty big PC enthusiast... I don't know if it's just me getting older or the technology being strong enough that an upgrade doesn't seem necessary anymore, but I have a brand new Ivybridge i7 3770k sitting at home on my desk and I haven't ever bothered to look at building a new PC around it.

I will eventually build on it. But part of me feels like my old Nehalem i5 machine is plenty of power for anything I am remotely interested in right now.

Intel hasn't given up on its innovation or delayed any of their designs for the next few generations as far as I know. The plans that were set years ago continue on and the client processor is still a high priority for the design teams. Maybe in the future the lack of competition will hurt innovation, but I wouldn't expect it to be noticeable for at least the next ~8 years. Price on the other hand... who knows. I would guess AMDs minor holdings will still keep prices reasonable. Intel also still needs to move enough units for the FABs and technology development to be worthwhile. This will help keep prices moderately low.

Clock speed is for sure important and yes currently it is a heavy cost to do it right now, but that is exactly why I said someone needs to get back to work on doing it. Most things server side would benefit from increased clock speeds, not to mention games. You don't think your games are capping your mhzs per core?

No, I don't... It's really silly to compare processors based on clock speed unless they are from the same family.

Instructions handled per cycle is a much more important measure of a processor. Each generation you will see a ~10% increase in performance per clock. That is to say, a 3.3GHz Sandybridge functions roughly equivalent to a 3.0GHz Ivybridge processor.

In addition to that increase im performance per cycle, more features being moved from the chipset onto the die is what sets one generation apart from the next. Looking for a massive increase in clock speed will just set you up for disappointment.

Who is comparing processors here? I sure as hell didn't post anything related to comparing CPUs Mhz, especially between architectures. So I'm not sure where you got that from. I agree it is really silly, that is why I didn't do it.

There are plenty of ways for you to completely saturate a single core though, the example that comes to my mind is Minecraft servers.

Hyperthreading was the way to go, for sure. Intel has just been all over it the past few years, while AMD prefers to add more cores.

In some cases, there isn't really a reason to upgrade year over year. My i7-870 at 4.0GHz would probably be fine playing most games and doing most things. That being said, I noticed almost NO difference going from the i5-760 to the i7-870. I noticed a huge difference going from the i7-870 to the i7-2600k, partly because I did overclock it to 5.0GHz.

There was a HUGE difference in gaming performance in the more demanding games that took advantage of multithreaded code. The 2600k was simply a beast. Now going from the 2600k to the 3700k isn't much of a gain. A gain, especially in efficiency and power usage, but not much from performance.

The reason clock speeds are not blazing fast is because they have not yet created ways to cool them efficiently. Sure, you can throw a decent air cooler or water cooling kit and crank the clocks to 4.5 GHz, but that isn't realistic for the everyday consumer.

I'm explaining why the listed clock speed is not important and not even remotely a primary concern among development teams working on future projects. You feel that it should be worked on, but that just is not a priority, nor do I see it becoming one in the near future. Companies will stretch their design, but not to what a lot of users seem to think is appropriate... Even if the lower clock speed results in better performance. Juicing up the clock is just extremely inefficient in terms of progress.

So cooling is definitely an issue for the higher voltages/speeds for sure. The die sizes shrink and because of that, the surface of the actual chip is more difficult to cool with contact...

Ivybridge is about as wide as the tip of your pinky finger and 3-4x that in height (depending on # of cores and graphics size). The packaging it sits in is MUCH larger than that, but the contact on the die itself is TINY in comparison to Sandybridge.

The other big issue is circuit degradation. Basically all of the circuits in the processor experience electromigration. This isn't a problem really with thicker metals, but as the process technology gets smaller and smaller, the metal connects are so tiny that this becomes a REAL problem.

The best way to think about it is like a river. As water runs across a riverbed, it wears down the rocks it is running across. The exact same thing happens with the metals. High current signals or constantly on DC currents especially experience this to a much greater and more noticeable degree.

While the processors are tested for these problems quite rigorously, defects happen in ANY process technology (think wire inconsistency and having thinner metal junctions in the circuitry) and the possibility of failure in these situations just goes up with the greater currents.

Companies then strive for the lowest possible failure rate while maintaining high enough performance gains to make up for the design effort... if that makes sense.