What I believe is tangible evidence of AMD's lack of competition

[VirtualLarry]

I don't know how many people remember this, but in the heydays of Core2Duo and socket 775, every few months, Intel would phase out their lowest-end chips, lower the price of the next-highest chips to that of those lower chips, and then introduce faster chips, at the higher price.

This just doesn't seem to be happening anymore. I cannot recall seeing any price roadmaps from Intel recently, with lower prices for their CPUs, and the introduction of new SKUs that are a speed-bin higher.

2500K? Stagnant.

3570K? Stagnant.

Is this what we have to look forward to with Haswell?

 

[Smoblikat]

Yes it is, if intel has no incentive to be faster why would they?

 

[TerryMathews]

I don't know how many people remember this, but in the heydays of Core2Duo and socket 775, every few months, Intel would phase out their lowest-end chips, lower the price of the next-highest chips to that of those lower chips, and then introduce faster chips, at the higher price.

This just doesn't seem to be happening anymore. I cannot recall seeing any price roadmaps from Intel recently, with lower prices for their CPUs, and the introduction of new SKUs that are a speed-bin higher.

2500K? Stagnant.

3570K? Stagnant.

Is this what we have to look forward to with Haswell?

Further back than that, same shit happened in the P2/P3 era before the introduction of the K7.

 

[mbreslin]

I agree! If it wasn't for the Athlon/K7, we would probably still have the Netburst architecture Pentium 4 on offer from Intel. We have AMD to thank for Intel getting off their butts and getting into both 64-bit CPUs and the core architecture.

 

[Sleepingforest]

There is/was a tiny bit of competition during Sandy Bridge. The i7-2600K was phased out in favor of the i7-2700K, I believe--though no price drop or other shifts happened.

 

[Plimogz]

Yes it is, if intel has no incentive to be faster why would they?

There's a position occasionally adopted by some which argues that Intel needs to improve regardless of AMD's performance in order to sell updated performance to their own past customers.

Anyway, yeah @OP. The introduction of the 2700K was typical of what you describe: it only supplanted the 2600K by the thinnest of margins, and instead of replacing the old part at the same price point, Intel saw fit to charge a little more for the extra performance. Which is fair enough -- in a sad way for enthusiasts -- but does contrast with the way that small incremental lineup updates would usually bump the previous SKU out of the price slot and offer slightly more performance for the same amount of money, as the months went by.

Also, I feel that it is no accident that Intel introduced the "K"s right at the top-end of AMD's performance range. Otherwise, once they'd clamped down bus OC, the unlocked Black Editions (and now FXs) would've offered better comparative value through overclocking at that point in the performance ladder (with pricing remaining the same). Hence Intel had reason to follow suite.

Then again, it could just be that Intel has its market sufficiently figured out that they know that they can upsell quads to people who would've otherwise settled for cheaper duals by locking the latter and adding a "K" to the former, regardless of what AMD is doing.

 

[mistercrabby]

Good discussion. Another angle is that Intel is the dominate corporate cpu. Corporate buyers like price stability and simpler product mix comodities for mainstream products. The OEMs know this and aren't pushing Intel to rock that boat.

 

[mistercrabby]

Another angle is that software architectures have plataued with respect to the performance requirements of standard hardware on properly sized servers. Windows and Linux are more efficient than ever. This applies to a large degree to desktops/laptops. Virtualization has made system resource utilization more and more effective, as have blades.

Intel is coasting until AMD gets bigger into the corporate market and pushes back on Intel's cost/performance curve. You know, like they used to do.

 

[ViRGE]

Does frequently releasing processors even make sense in the age of turbo though? Both AMD and Intel have gone to great lengths to min-max their products through turbo. There's not much room for turning up the clocks, at least not if you hold at the same TDPs and voltages.

 

[Roland00Address]

They still do this but they do so in laptops not desktops.

Every 3/4 months there is a new version of each i3, i5, and often i7 processors.

Sandybridge
i3 2310m at 2.1 ghz, Feb-11
i3 2330m at 2.2 ghz, Jun-11
i3 2350m at 2.3 ghz, Oct-11
i3 2370m at 2.4 ghz, Jan-12
Ivybridge
i3 3110m at 2.4 ghz, Jun-12
i3 3120m at 2.5 ghz, Sept-12
i3 3130m at 2.6 ghz, Jan 13

So in 2 years we have gotten a free improvement of 24% in mhz bumps, plus the small 5 to 10% performance increase from going from sandybridge to ivybridge in cpu tasks, and the much better intel hd 3000 vs 4000.

--------------------

Now one reason Intel is not doing this in desktop is the desktop market is much more mature; mature being a preoperative word, other pejorative words that could accurately describe the situation is stagnant or perhaps the word stable.

The second reason is, each bump in mhz speed increases power consumption unless the voltage is decreased to compensate. Note the following chart

Note that after 3.0 ghz the increase in mhz and thus performance does not have a linear ratio with the corresponding increase in power consumption. After 3.5 and 4.0 ghz the power consumption ratio goes much higher

 

[Plimogz]

Wouldn't binning and process improvements still enable otherwise identical but slightly faster SKUs in time, turbo or not?

 

[svenge]

Wouldn't binning and process improvements still enable otherwise identical but slightly faster SKUs in time, turbo or not?

Yes. That's why Intel is releasing some new SKUs in the near future that are 100MHz faster than their predecessors. Of course they're all in the Pentium and i3 lines, but that's where the competition vs AMD currently lies.

 

[Idontcare]

Wouldn't binning and process improvements still enable otherwise identical but slightly faster SKUs in time, turbo or not?

The traditional means by which faster sku's would be generated as a node matured is no longer practical or relevant these days.

See back in the day, the days the OP is thinking of, it was common to release a node from R&D to manufacturing that was still quite immature and not meeting its targeted parametric (electrical values).

So there was at least a year's worth of more work to be done by the production team to tweak and optimize the process node such that it eventually finally hit its parametrics and the chips that had been designed to operate at 3GHz (or whatever target there was) actually start functioning at that speed and within the intended power consumption envelope.

As outsiders we saw this progress in the form of steady releases of ever slightly faster CPUs, 100-200MHz bumps in speed at a time.

That approach to process node development went away with 45nm for just about everyone but AMD (GloFo). AMD was still scrambling and scrapping to get their nodes out regardless how immature they were at the time (which created the Llano shortfall and so on at 32nm).

Nowadays the modus operandi is to keep the process node with the R&D group and not allow its release to production until it hits its parametrics. That leaves the production guys with less to worry about, leaving them to focus on doing their actual job which is to do things that lower production costs (rather than trying to do things that increase the ASP of the chips). They focus on reducing cycle time, increasing functional yields, lowering emissions to decrease permitting costs and so on.

Because of this the chips that come out of the fab in the first quarter are already hitting their parametrics. Instead of shipping 3GHz now and slowly creeping up to launching 3.5GHz chips next year the fabs can launch with 3.5GHz chips out of the gate at time zero.

What this looks like to the outsider is the company is milking consumer an not releasing faster products over time when it really has nothing to do with that. It is just a different way of managing the transition of what is called "productization" of the new node as it is handed off from R&D to manufacturing.

It is done this way now because it is actually better in the big picture as it simplifies the complexities of an already immensly complex production situation. When you have to spend a few billion dollars just to develop a new node, and another few billion dollars just to develop a functioning chip for that new node, minimizing complexity is critical to your economic survival.

 

[MaxPayne63]

Didn't it take over 3 years to go from Prescott to Conroe?

I'm pretty sure every (consumer desktop) CPU since Nehalem has been released 12-15mos after the previous codename.

e: And Conroe arrived after nearly seven years of netburst happy fun days.

 

[ViRGE]

Didn't it take over 3 years to go from Prescott to Conroe?

I'm pretty sure every (consumer desktop) CPU since Nehalem has been released 12-15mos after the previous codename.

e: And Conroe arrived after nearly seven years of netburst happy fun days.

I think you're forgetting about Cedar Mill, the 65nm P4. That was released in January of 2006 (about 2 years after Prescott).

 

[VirtualLarry]

I think you're forgetting about Cedar Mill, the 65nm P4. That was released in January of 2006 (about 2 years after Prescott).

When did tick-tock start? Cedar Mill, or Conroe?

 

[meloz]

Is this what we have to look forward to with Haswell?

Yes, more of the same on the CPU side. iGPU continues to improve though, as does the overall system energy efficiency.

I would disagree with our assesment that the stagnation in outright performance (as demonstrated by 2500K and 3570K) indicate AMD's uncompetitiveness. As other posters have alluded in past, it appears Intel have truly wrung out all the performance they can from x86-64, and only way they can gain any new absolute performance increase is through new instructions (AVX2 etc). This is not taking into account improvement in performance / watt.

IMO a better indicator of AMD's lack of competitiveness is the price of the CPUs and chipsets of the two companies. Some delusional people still contend that certain AMD processors are performance competitive with Intel offerings. What they fail to take into consideration is the stark price difference between these supposedly 'equivalent' products.

The premium Intel charge for their CPUs , IMO, is far more indicative of their domination and AMD's utter failure.

 

[Charles Kozierok]

The tick-tock business started with Conroe and the Core 2 Duo.

 

[Torn Mind]

Yes, more of the same on the CPU side. iGPU continues to improve though, as does the overall system energy efficiency.

I would disagree with our assesment that the stagnation in outright performance (as demonstrated by 2500K and 3570K) indicate AMD's uncompetitiveness. As other posters have alluded in past, it appears Intel have truly wrung out all the performance they can from x86-64, and only way they can gain any new absolute performance increase is through new instructions (AVX2 etc). This is not taking into account improvement in performance / watt.

IMO a better indicator of AMD's lack of competitiveness is the price of the CPUs and chipsets of the two companies. Some delusional people still contend that certain AMD processors are performance competitive with Intel offerings. What they fail to take into consideration is the stark price difference between these supposedly 'equivalent' products.

The premium Intel charge for their CPUs , IMO, is far more indicative of their domination and AMD's utter failure.

IMO, the strongest evidence of AMD's non-competitiveness is the difference in the MSRPs of the FX-x1xxs and the FX-x3xxs, such as the FX-8150's $270 MSRP compared to the FX-8350's $195 MSRP.

 

[2is]

There's a position occasionally adopted by some which argues that Intel needs to improve regardless of AMD's performance in order to sell updated performance to their own past customers.

Which is true. But with competition they'd need to do it even better/faster.

 

[VirtualLarry]

There's a position occasionally adopted by some which argues that Intel needs to improve regardless of AMD's performance in order to sell updated performance to their own past customers.

As outsiders we saw this progress in the form of steady releases of ever slightly faster CPUs, 100-200MHz bumps in speed at a time.

Because of this the chips that come out of the fab in the first quarter are already hitting their parametrics. Instead of shipping 3GHz now and slowly creeping up to launching 3.5GHz chips next year the fabs can launch with 3.5GHz chips out of the gate at time zero.

What this looks like to the outsider is the company is milking consumer an not releasing faster products over time when it really has nothing to do with that.

But isn't every speed bump increase, a marketing opportunity? By releasing only a few CPUs at "max speed" for the process, instead of milking the market with a number of incremental increases over the lifetime of that CPU's design and process node, doesn't Intel lose all of that CPU revenue that they might have received, from "serial upgraders"? And even if not for them, for the many more opportunities to score an upgraded CPU sale from Joe six-pack along the way?

Or are there so few people that actually bother upgrading their CPUs, that this line of thinking is pointless?

Edit: Or is the cost savings from the R&D and Process engineering savings, from waiting until the process is hitting the parametrics for the design, etc., before producing the CPUs, more than the possible sales profits from serial upgraders?

 

[2is]

I think most people just overclock. The "serial upgraders" market isn't one Intel is worried about. For example, I doubt there were enough people that upgraded their 2600k to a 2700k for Intel to care. Most builders with a 2600k were already a step ahead of a 2700k and there wasn't much compelling evidence showing the reached a higher OC either.

 

[escrow4]

CPU performance isn't important anymore. Its all mobile and thin and light - the whole package counts not raw grunt. I upgraded from a Phenom II to Ivy Bridge. Night and day. Now I'll think about upgrading around Broadwell, if that. In 2013, massive performance boosts just are not what Intel is targeting. AMD has little effect on Intel CPU wise anyway. There is no competition anymore, just an alternative.

 

[Brekyrself]

CPU performance isn't important anymore. Its all mobile and thin and light - the whole package counts not raw grunt. I upgraded from a Phenom II to Ivy Bridge. Night and day. Now I'll think about upgrading around Broadwell, if that. In 2013, massive performance boosts just are not what Intel is targeting. AMD has little effect on Intel CPU wise anyway. There is no competition anymore, just an alternative.

This. Also, what percent of the market share actually needs anything faster? Besides enthusiasts, anything newer then a p3 can browse the internet and go on facebook....

 

[StrangerGuy]

This. Also, what percent of the market share actually needs anything faster? Besides enthusiasts, anything newer then a p3 can browse the internet and go on facebook....

There is CPU competition and more intense than ever...Except they aren't playing it with Intel's rules.