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

[Pilum]

Ah, there goes the AMD sob story again.

Yes, but this time with the Soviet Union and the Roman Empire thrown in. This may get interesting yet.

 

[dagamer34]

I agree with you, but would argue that by your own definition, intel has failed to compete successfully with itself, at least in the desktop market. Can you really give me a good reason why one should upgrade from sandy bridge to ivy or haswell?

Mobile may be another story, hopefully it will, but that is yet to be seen.

No normal person upgrades their computer on 2 year product cycles. This isn't the 90s anymore where 2 years meant a sea change in performance.

I'd say Intel needs to consistently compete with products they released 4 years ago to keep moving forward. Trying to cater to enthusiasts who want 20-30% better performance compared to their overclocked 4.8Ghz Sandy Bridge chips when they ship them at 3.4Ghz is just... never going to happen.

 

[vampirr]

http://are.berkeley.edu/~sberto/AMDIntel.pdf

http://venturebeat.com/2009/09/21/e...-of-evidence-against-intel-in-antitrust-case/

http://europa.eu/rapid/press-release_MEMO-09-400_en.htm?locale=en

http://www.theregister.co.uk/2008/05/15/intel_amd_lawsuit_internal_interviews_order/

http://news.cnet.com/8301-30685_3-10390568-264.html

The sources are reliable and remember since you are surely from USA, your goverment and your laws confirmed what Intel did. Deal with it...


OK, warnings over, vacation time.
Markfw900
Anandtech Moderator

 

[Homeles]

That's nice. Your ridiculous argument also completely ignores the incompetencies of AMD's management. Ever heard of cherry picking?

 

[mrmt]

The sources are reliable and remember since you are surely from USA, your goverment and your laws confirmed what Intel did. Deal with it...

So what?

 

[frozentundra123456]

http://are.berkeley.edu/~sberto/AMDIntel.pdf

http://venturebeat.com/2009/09/21/e...-of-evidence-against-intel-in-antitrust-case/

http://europa.eu/rapid/press-release_MEMO-09-400_en.htm?locale=en

http://www.theregister.co.uk/2008/05/15/intel_amd_lawsuit_internal_interviews_order/

http://news.cnet.com/8301-30685_3-10390568-264.html

The sources are reliable and remember since you are surely from USA, your goverment and your laws confirmed what Intel did. Deal with it...

It was dealt with, get over it.

 

[Idontcare]

I agree with you, but would argue that by your own definition, intel has failed to compete successfully with itself, at least in the desktop market. Can you really give me a good reason why one should upgrade from sandy bridge to ivy or haswell?

Mobile may be another story, hopefully it will, but that is yet to be seen.

My "2 yr" number was merely a reference point, I wasn't speaking explicitly about the timeframe.

Whether it is 2yrs ago, 3, or 4, or 5...whatever the upgrade cycle tends to be that defines the bulk of the TAM is what matters. And 80% of those chips sitting in desktops and laptops have an Intel logo on them.

If I had to guess I'd say Intel is probably targeting a 4-5 yr upgrade cycle for the bulk of its upgrades.

But that is a rolling 4-5yr cycle, so Haswell doesn't need to be a compelling upgrade to SB, just a compelling upgrade to Conroe/kentsfield or Yorkfield/Penryn for Intel to be able to mine enough billions out of the upgrade TAM to keep the revenues growing.

 

[svenge]

Deal with it...

When will you deal with the fact that you're mindlessly shilling for a moribund company with dire financial prospects and grossly inferior products?

 

[IntelUser2000]

I agree with you, but would argue that by your own definition, intel has failed to compete successfully with itself, at least in the desktop market. Can you really give me a good reason why one should upgrade from sandy bridge to ivy or haswell?

Mobile may be another story, hopefully it will, but that is yet to be seen.

Let me say first that even mobile claims might be exaggerated, and its only proven when we see it.

But in general, they might have anticipated this. This might be why they are significantly reducing the ramp of Desktop Haswell(ramp means how long it takes for a product to take over a certain market).

Also the fact that the Desktop Haswell is coming in June rather than earlier like May or April is a sign. My unofficial rule says there's not something completely satisfying with the chip when the product is delayed. Remember Ivy Bridge? With delay, came disappointment.

I wouldn't be surprised that we would see average 10% gain in the final revision, few % up from the current 7-8%, but nothing enormously beyond that. It also seems similar to Nehalem in that area that advances were based on memory bandwidth and SMT, and shined in a specific area called server. Haswell also needs specialization like AVX2 and TXT, and may only shine in specific areas(like mobile, but again to be seen).

 

[loccothan]

Thats why i choose AMD !!!! Intel will Pay for its Backstabing Politics against AMD

http://www.youtube.com/watch?v=4et7kDGSRfc

Here u have some truth about so called Best Intel, of course the truth is even Phenom II is better in calculations than i7 3770k !!!!!
And remember the consoles launch PS4 and xBx 720 !!! will be AMD chip inside !!
That means the optimalisation will be for AMD at least for 8 years !!!!
New 3DsMax 13, the Patch for ADOBE and lot more, of course Windows 8 is Great for FX processors !!

 

[Torn Mind]

Yes, but this time with the Soviet Union and the Roman Empire thrown in. This may get interesting yet.

Roman culture did not value innovation and were a rather practical people.

The Soviet Union was quite a miserable place to live even with it's iron fist-like enforcement of order.

If anything, the narrative of downfall would be more applicable to Intel perhaps when they get their "bad CEO", in which the big guy gets burnt out and dies because it didn't manage their resources well to stop the enemy.

 

[taltamir]

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.

For that to occur intel must first saturate the market. And that isn't happening soon.

 

[Idontcare]

For that to occur intel must first saturate the market. And that isn't happening soon.

When Intel is seeing declining shipments, meaning demand is no longer meeting supply, wouldn't you consider that to be a "saturated market"?

I don't think Intel has ever ran "at capacity" meaning they have always saturated the market with all the Intel product the market cares to buy at Intel's prices.

Maybe you mean something different by "saturate"? Or maybe I just don't understand?

 

[shady28]

do you believe that AMD is capable of doing so? my impression is that AMD is really hurting and not as technologically capable.
...

Agree with you 100%. And for comparison, Intel's R&D budget it 800% the size of AMDs.

Given that simple fact, I'd say AMD is doing remarkably well vs Intel. If AMD were to outstrip Intel in performance and/or performance/watt, it would be a clear sign of a vastly bloated and inefficient Intel.

And if Steamroller gives us 20% IPC boost from Piledriver, they will be getting mighty close to Intel on single thread.

 

[IntelUser2000]

Agree with you 100%. And for comparison, Intel's R&D budget it 800% the size of AMDs.

That's overblown though. They have LOT of research projects going on, and they have their fabs as well. Relatively small amount is entirely CPUs.

 

[Markfw]

FYI vampirr has a few days off, maybe weeks. He has ignored my repeated warnings.I will not put up with his crap or any like it.

Official notice, I am pissed

 

[shady28]

That's overblown though. They have LOT of research projects going on, and they have their fabs as well. Relatively small amount is entirely CPUs.

AMD's numbers also include things like developing the new Radeon cores as well though, and they've branched a bit with APUs.

But yes, Intel is heavy into Fab research. After all, process tech shrinks the die and they can put more CPUs on one wafer. Heck, top line i5's are half the size of an FX-8350 already.

This is what we're seeing though, and if you read it carefully it shows that we're at or near a point where process / fab tech won't matter :

http://www.extremetech.com/computin...rom-one-core-to-many-and-why-were-still-stuck

"Given the low performance returns… adding more cores will not provide sufficient benefit to justify continued process scaling. Given the time-frame of this problem and its scale, radical or even incremental ideas simply cannot be developed along typical academic research and industry product cycles… A new driver of transistor utility must be found, or the economics of process scaling will break and Moore’s Law will end well before we hit final manufacturing limits"

 

[Idontcare]

That's overblown though. They have LOT of research projects going on, and they have their fabs as well. Relatively small amount is entirely CPUs.

IC design does take up the bulk of R&D budgets though.

Look at TSMC's budget to get a feel for what it costs to develop leading edge (2nd only to Intel) process nodes. It takes a billion or so per year.

Next look at AMD's budget - that 1.4B is nearly entirely IC design expense. As is nvidia's, Marvell's. Heck look at how much Qualcomm and Broadcom spend on IC design every year.

Now take Intel's 8+ billion R&D into consideration. Maybe $1.5B of that is process node. R&D budgets don't include fabs, that is a seperate budget called capex. So you can estimate Intel spent $7B of their R&D budget on IC designs (and that was just for 2011 alone ).

How much of that went into mainstream x86 vs. Atom vs. Itanium vs. SSD and other misc projects? That we can't answer, but the performance of their products (in terms of IPC) would suggest they are investing considerably more money into x86 development than AMD.

 

[JimmiG]

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?

I don't think the lack of competition is the only reason. If it was *easy* to make a CPU that's significantly faster than the 2500K, AMD would have released one.

-There's a clock speed wall at around 4 GHz. Intel hit 3.8 GHz around 2004, and no CPU since then has been clocked much higher.

-Also, Intel is approaching an IPC wall. All the low-hanging fruit has been picked, making it harder and harder to squeeze out more performance per clock, per core.

-There's a limit to how many cores you can utilize. Some code just has to be executed serially rather than in parallel. Before you can do A+B, you have to figure out what A and B actually are. This seems to top out at 4-6 cores for typical desktop use/gaming, a little more for some highly specialized tasks.

AMD and Intel are both running out of options, IMO. It's just that Intel got a bit farther before this happened. I guess that's why everyone is going crazy about smart phones and tablets. I think desktop CPUs will remain stagnant while mobile SOCs creep closer to the same performance level as desktop CPUs and GPUs. This is already happening.

 

[djgandy]

Also Intel processors are as cheap as they have ever been. You can buy the top of the line desktop CPU for ~$300 (excluding Socket 2011). Back in 2006 that only got you a Q6600. Of course that was lately replaced by a faster CPU that used to cost $572 but was bumped down into the $266 slot. But both chips were the same.

If you look at it like this, Intel was actually sandbagging back in 2006 because C2D was so good. They held back on the clocks so that they could sell the same product for more money over time.

 

[taltamir]

When Intel is seeing declining shipments, meaning demand is no longer meeting supply, wouldn't you consider that to be a "saturated market"?

I don't think Intel has ever ran "at capacity" meaning they have always saturated the market with all the Intel product the market cares to buy at Intel's prices.

Maybe you mean something different by "saturate"? Or maybe I just don't understand?

To be accurate, market saturation is a percentage rather then a yes/no bool. I said "saturated market" and that was misleading. I meant very high saturation, closing in on 100%.
At full 100% saturation shipments will not just decline, they will drop to near 0% (where only people buy new products are those whose existing product broke)
And when I said saturation I meant "own current product or equivalent" rather then "willingness to buy at current price".

Regardless of what the price is, I currently own what intel is selling. Any sales pitch by them, discount, or whatever will be met with "yes its a wonderful product, I already own it".
I do not need/benefit from buying more of them just to have a dozen CPUs sitting on my shelf. The portion of the market that I represent is fully saturated regardless of price because I have everything I could possibly want that they are currently manufacturing. Of course if you make something cheap enough people might buy it anyways (eg, I could strip it for metals to sell to recycling plants or repurpose it as a paperweight) so I meant within reason.

The only way for intel to sell me a new CPU is for them to design one that is an actual improvement over what they already sold me. I would say that the vast majority of people are not in that position. There are the developing markets and population increase to contend with of course, but even in mature markets like the USA many people own ancient computers and are too cheap to upgrade. My dad is using an ancient slow computer and he might be enticed to buy... Although it seems to me intel is instead playing chicken with those people. Rather then producing an actual upgrade for people like me to buy, they are constantly lowering the multipliers to keep performance about the same and waiting for people like my dad to break and buy.

Of course I am grossly oversimplifying. Intel also drips us slight improvement over time.

 

[StrangerGuy]

To be accurate, market saturation is a percentage rather then a yes/no bool. I said "saturated market" and that was misleading. I meant very high saturation, closing in on 100%.
At full 100% saturation shipments will not just decline, they will drop to near 0% (where only people buy new products are those whose existing product broke)
And when I said saturation I meant "own current product or equivalent" rather then "willingness to buy at current price".

Regardless of what the price is, I currently own what intel is selling. Any sales pitch by them, discount, or whatever will be met with "yes its a wonderful product, I already own it".
I do not need/benefit from buying more of them just to have a dozen CPUs sitting on my shelf. The portion of the market that I represent is fully saturated regardless of price because I have everything I could possibly want that they are currently manufacturing. Of course if you make something cheap enough people might buy it anyways (eg, I could strip it for metals to sell to recycling plants or repurpose it as a paperweight) so I meant within reason.

The only way for intel to sell me a new CPU is for them to design one that is an actual improvement over what they already sold me. I would say that the vast majority of people are not in that position. There are the developing markets and population increase to contend with of course, but even in mature markets like the USA many people own ancient computers and are too cheap to upgrade. My dad is using an ancient slow computer and he might be enticed to buy... Although it seems to me intel is instead playing chicken with those people. Rather then producing an actual upgrade for people like me to buy, they are constantly lowering the multipliers to keep performance about the same and waiting for people like my dad to break and buy.

Of course I am grossly oversimplifying. Intel also drips us slight improvement over time.

As a gamer I'm still using a nearly 4 year old HD5850, let alone a huge amount of people with C2Ds and better whose most demanding app are Flash webgames.

Developing markets? Pffft. There will be an entire generation of people in those whose computing experience revolves around an ARM tablet that cost less than what Intel charges for a single Pentium.

 

[geniusloci]

Features matter more than performance now. I have dual thunderbolts on my board because even though I'm not using them right now, I might in 6 months or a year. I don't expect the 2600k I have on this board to need to be upgraded for several generations. The software just isn't there to use it.

I think we're really headed into that situation where intel collapses the motherboard market and sells the average person NUCs. Dear Intel, please give me lots and lots of ports. I hate running out of usb/thunderbolt/whatever.

 

[taltamir]

As a gamer I'm still using a nearly 4 year old HD5850, let alone a huge amount of people with C2Ds and better whose most demanding app are Flash webgames.

Did you have a point?

Developing markets? Pffft. There will be an entire generation of people in those whose computing experience revolves around an ARM tablet that cost less than what Intel charges for a single Pentium.

Developing markets does not mean subsistence farmers in africa.
The world as a whole is becoming more industrialized and there are places which have just enough money to buy.

Features matter more than performance now. I have dual thunderbolts on my board because even though I'm not using them right now, I might in 6 months or a year.

Thunderbolt's only feature is that its fast. aka, performance.

 

[geniusloci]

Thunderbolt's only feature is that its fast. aka, performance.

I should have spelled out that I was obviously referring to platform features or extras as opposed to cpu performance, but I would think the guy with 13,000 posts would have grasped such a thing.