AMD and Intel

[MDme]

Do you think AMD has "spies" on Intel CPU teams or vice versa? I mean, do you think Intel knows actual performance characteristics of K8L? or will AMD know the performance of Penryn (if it is already running)? I mean for C2D, Intel released public benchmarks 6 months before they launched so AMD would have known by then. But what about the current situation? Does intel have inside intelligence knowing K8L performance?

* This is not and should not be a flame-bait thread. This is intended to discuss if either company has inside guys.

 

[tersome]

I'm guessing that it's probably illegal to have industrial spies, and Intel wouldn't be stupid enough to risk getting caught.

I have no idea though.

 

[SickBeast]

I'm sure they follow each other's patents and press releases.

There is also probably a great deal of crossover among past employees at both companies.

There are also probably a great deal of engineers who are friends who work at different companies.

I doubt they pay spies tho. The money is better spent on research.

 

[MDme]

it is true that money is better spent on research but the money you'd spend for paying a "spy" would not compare to the R&D budget to begin with. I also agree with you that there are friends each work in the other company. They also have a cross-licensing agreement anyway. It is just intriguing to think if they are aware of what the others chip/design can do for example: IF Intel had an idea how K8L would perform, it would probably affect it's execution of it's roadmaps (i.e. if K8L is a beast, would they accelerate penryn? or maybe delay it to tweak it and make it better than K8L? or if K8L was not significantly better, would they not feel pressure to release penryn since C2D has good clock headroom?).

On another note, does AMD's lack of a benchmark (so close to actual launch of K8L Q2 07) mean AMD is trying to surprise Intel (if they didn't have spies) or does it mean that K8L simply is not up to snuff vs C2D/penryn?

Just pondering...

 

[SickBeast]

Originally posted by: MDme
On another note, does AMD's lack of a benchmark (so close to actual launch of K8L Q2 07) mean AMD is trying to surprise Intel (if they didn't have spies) or does it mean that K8L simply is not up to snuff vs C2D/penryn?

Just pondering...

I don't think AMD ever releases benchmarks prior to a CPU launch (neither does intel).

They did, however, state that their chip would be 40% quicker than the intel chips. How they are making that comparison, I don't know.

In response to your other stuff: I don't think intel/AMD worry too much about each other's product designs until they are released. Even then, they tend to take very unique approaches to CPU design. Look at the IMC of the A64 - intel still has yet to implement one, and got around it quite elegantly with the C2D memory controller.

IMO it will be another 5-10 years before intel has to seriously worry about AMD. Either that or AMD will go broke.

 

[MDme]

Originally posted by: SickBeast

Originally posted by: MDme
On another note, does AMD's lack of a benchmark (so close to actual launch of K8L Q2 07) mean AMD is trying to surprise Intel (if they didn't have spies) or does it mean that K8L simply is not up to snuff vs C2D/penryn?

Just pondering...

I don't think AMD ever releases benchmarks prior to a CPU launch (neither does intel).

They did, however, state that their chip would be 40% quicker than the intel chips. How they are making that comparison, I don't know.

In response to your other stuff: I don't think intel/AMD worry too much about each other's product designs until they are released. Even then, they tend to take very unique approaches to CPU design. Look at the IMC of the A64 - intel still has yet to implement one, and got around it quite elegantly with the C2D memory controller.

IMO it will be another 5-10 years before intel has to seriously worry about AMD. Either that or AMD will go broke.

Good points but:
1) Intel did release C2D benches against an OC'd FX 6 months prior to launch.
2) I think intel was worried about AMD (i don't know it it was "seriously worried") that's why we have C2D chips now. That's why we have this brutal price war now.

 

[SickBeast]

Originally posted by: MDme

Originally posted by: SickBeast

Originally posted by: MDme
On another note, does AMD's lack of a benchmark (so close to actual launch of K8L Q2 07) mean AMD is trying to surprise Intel (if they didn't have spies) or does it mean that K8L simply is not up to snuff vs C2D/penryn?

Just pondering...

I don't think AMD ever releases benchmarks prior to a CPU launch (neither does intel).

They did, however, state that their chip would be 40% quicker than the intel chips. How they are making that comparison, I don't know.

In response to your other stuff: I don't think intel/AMD worry too much about each other's product designs until they are released. Even then, they tend to take very unique approaches to CPU design. Look at the IMC of the A64 - intel still has yet to implement one, and got around it quite elegantly with the C2D memory controller.

IMO it will be another 5-10 years before intel has to seriously worry about AMD. Either that or AMD will go broke.

Good points but:
1) Intel did release C2D benches against an OC'd FX 6 months prior to launch.
2) I think intel was worried about AMD (i don't know it it was "seriously worried") that's why we have C2D chips now. That's why we have this brutal price war now.

Intel's prices remind me of the P2 days. They are dominating and they know it. They can charge $200 for an entry level CPU which is a very good sign for them. AMD just cut $15 off a 3800+ CPU. Not drastic cuts by any means.

It's not a "war" right now. Intel owns the high end, and AMD owns the low end.

Intel came up with C2D because the P4 was an engineering failure. They did not anticipate the termal wall they hit, nor did they seem to realize the inefficiency of working with a massive pipeline on a CPU.

Really to be honest it seems like intel ignores AMD completely, almost out of pride.

 

[MDme]

Originally posted by: SickBeast

Originally posted by: MDme

Originally posted by: SickBeast

Originally posted by: MDme
On another note, does AMD's lack of a benchmark (so close to actual launch of K8L Q2 07) mean AMD is trying to surprise Intel (if they didn't have spies) or does it mean that K8L simply is not up to snuff vs C2D/penryn?

Just pondering...

I don't think AMD ever releases benchmarks prior to a CPU launch (neither does intel).

They did, however, state that their chip would be 40% quicker than the intel chips. How they are making that comparison, I don't know.

In response to your other stuff: I don't think intel/AMD worry too much about each other's product designs until they are released. Even then, they tend to take very unique approaches to CPU design. Look at the IMC of the A64 - intel still has yet to implement one, and got around it quite elegantly with the C2D memory controller.

IMO it will be another 5-10 years before intel has to seriously worry about AMD. Either that or AMD will go broke.

Good points but:
1) Intel did release C2D benches against an OC'd FX 6 months prior to launch.
2) I think intel was worried about AMD (i don't know it it was "seriously worried") that's why we have C2D chips now. That's why we have this brutal price war now.

Intel's prices remind me of the P2 days. They are dominating and they know it. They can charge $200 for an entry level CPU which is a very good sign for them. AMD just cut $15 off a 3800+ CPU. Not drastic cuts by any means.

It's not a "war" right now. Intel owns the high end, and AMD owns the low end.

Intel came up with C2D because the P4 was an engineering failure. They did not anticipate the termal wall they hit, nor did they seem to realize the inefficiency of working with a massive pipeline on a CPU.

Really to be honest it seems like intel ignores AMD completely, almost out of pride.

Not when they benchmarked C2D against A64FX (which was the first time they actually benchmarked a system against a "competitor's" cpu - they even Oc'd it.

 

[SickBeast]

I guess losing the performance lead for over a year woke intel up.

They're still arrogant tho.

 

[zsdersw]

The P4 was not an engineering failure. Engineering failures are things that don't work. The P4 worked just fine. The fact that it didn't reach the speeds they thought it would doesn't make it an engineering failure.

How is it an engineering failure when a Cedar Mill P4 makes it to 8GHz?

 

[gOJDO]

@NDme
Yes, they have spies. But no one can know how a future archietcture CPU will perform exactly. So, knowledgable engineers are estimating the perofmrance, based on the known informations about the features included in that architecture. C2D benchmarks were released 6 months before its launch because Intel knew that AMD have nothing to compete against. It was a good marketing. They told people to hold on and wait for their new CPU, because it rocks. I was waiting for C2D as well.
Now the situation is different. Both will bring new CPUs on the table, but are a bit quiet. I think that once K8L ES benchmars appear, there would be Penryn ES benchmarks too.

@SickBeast
Those statements from Randy Allen came after bad 2K6 Q4 earning report. He stated that Barcelona will offer 40% more performance over Clovertown on variety of workloads, based on SPEC_fp. He also stated that it will be 10%-15% faster in TPC-C. Both SPEC_fp and TPC-C are very bandwidth dependend benchmarks and are not representing the real world. IMO, K8L will be a FPU monster, it will be slower in ALU performance compared to Core2 and their SSE performance will be roughly same. Clock, for clock, one will be marginaly faster than the other. Penryn will dominate K8L because of its high freqfency.

@zsdersw
The P4 was the greatest disaster in Intel's history. It is also Intel's most unefficient(power, money and clock wise) architecture in x86 history. It costed Intel 3 years, a lot of invested money, lost market share and lost nerevers..
Cedar Mill P4 reached 8GHz with LN2, only for a few minutes, just for a CPUz validation. For this were needed a lot of people and expencive equipment. It is not a sucess! Intel was designing Netburst CPUs to reach such freqfencies using normal air cooling, but they never released Netburst space heater clocked higher than 3.8GHz.

 

[zsdersw]

Incorrect. The P4 was not a disaster. Northwood was a resounding success. Prescott was not, neither was Williamette, but Presler was okay and a good way to end the Netburst architecture.

 

[nanaki333]

if anyone knew of spies, there wouldn't be any

 

[gOJDO]

Originally posted by: zsdersw
Incorrect. The P4 was not a disaster. Northwood was a resounding success. Prescott was not, neither was Williamette, but Presler was okay and a good way to end the Netburst architecture.

I don't agree. P4 CPUs were very expencive and very performance/price uncompetetive. Although NorthwoodC outperformed the K7, it was very expencive and at same price level there wasn't any Northwood that coult hold its own against K7. In the beginign of 2002 I purchased two systems with all same components, except the mainboard, RAM and CPU:
1. Asus P4T-E, Intel 850i, AGP x6, RDRAM-800
512MB (2x256MB) Samsung Rambus 800MHz
Intel P4(Northwood) 1.8GHz 512kB L2 FSB-400
2. MSI K7T266Pro-2, VIA KT266A, AGP x8, DDR266
512MB (2x256MB) Kingston DDR-266 CL2
AthlonXP 1800+ 1.53GHz 256kB L2 FSB-266
Although the Intel system was $2000 and the AMD system was $1700, the 1800+ outperformed the Northwood in most real-life apps, especialy in gaming.
And not only Northwood, but all P4 were very performance/watt unefective, especialy the Prescott2M which is the greatest waste of silicon of all P4.
Pentium D was a follow-up of the Netburst disaster. All Smithfield CPUs got spanked by any Athlon64 X2 and dualcore Opteron CPUs. Presler was a little more effective, but again it was just an expencive space heater with relatively poor performance.
If Intel skiped the P4 and instead puted all their efforts into the P3 develeopment we would have got a Core2 CPUs three years ago.
With the discountinuing of the Netburst, and going back to P3 derivate Yonah, Intel just admited their mistake. IMO they should have killed Netburst with Northwood, after K8 arrived.

 

[zsdersw]

Core 2 Duo outperforms K8 right now.. but does that make K8 a "disaster"? No, it doesn't. Northwood was very price competitive when it wasn't limited to RDRAM.

 

[MDme]

@gOJDO

They don't need spies once the "specs" are released. but what about spies for let's say... sneaking out an ES to bench and get "realworld" estimates of performance. wouldn't that enable the company improve their design if it wasn't enough or maybe adjust clock bins/launch frequencies to better compete?

 

[BernardP]

AMD spying on Intel would be like

Pepsi spying on Coca-Cola

The clincher: The alleged plans were foiled after Pepsi, based in Purchase, N.Y., warned Coca-Cola

Look at the storm a little boardroom infighting has created at HP. It seems the risks related to getting caught would keep both AMD and Intel from engaging in spying.

 

[Regs]

I believe there a lot of smart businessmen and market analysts that can predict where a company is headed or what market segments they are going to be strong in.

Innovations and product line up no one will ever know until it hits the consumer base.

 

[BrownTown]

They probably don't spy, but I'm sure they are probably following what the other is doing VERY closely, it would not be hard to at least get a pretty good idea of what is comming. Exact numbers maybe not, but certianly they have a good picture of what is likely comming.

 

[Viditor]

Originally posted by: MDme
Do you think AMD has "spies" on Intel CPU teams or vice versa? I mean, do you think Intel knows actual performance characteristics of K8L? or will AMD know the performance of Penryn (if it is already running)? I mean for C2D, Intel released public benchmarks 6 months before they launched so AMD would have known by then. But what about the current situation? Does intel have inside intelligence knowing K8L performance?

* This is not and should not be a flame-bait thread. This is intended to discuss if either company has inside guys.

I agree with the consensus, there's likely no real "spying" going on.
However, remember that AMD and Intel have a huge cross-licensing agreement...and patents can tell you an awful lot about what the other guy is doing.

 

[myocardia]

Originally posted by: zsdersw
Core 2 Duo outperforms K8 right now.. but does that make K8 a "disaster"? No, it doesn't. Northwood was very price competitive when it wasn't limited to RDRAM.

Feel free to repeat it as many times as you like, it still doesn't make you right. Those of us who know what we're talking about know that Netburst was a disaster after Northwood, and should have been ended with the Northwood, which was the only successful time period of Netburst. Unless you think that continuing to make P4's, which were good competition for the K7's, while your competitor has the K8 makes good sense?:laugh: Would you not ridicule AMD for sticking with Skt. 939 for another 3 years, like Intel did with the P4, even though the C2D has been released?

Concerning spies, I think there probably are a few. I'll edit this post to include their names and titles, just as soon as I they are made publicly available.

 

[Gamingphreek]

Originally posted by: myocardia

Originally posted by: zsdersw
Core 2 Duo outperforms K8 right now.. but does that make K8 a "disaster"? No, it doesn't. Northwood was very price competitive when it wasn't limited to RDRAM.

Feel free to repeat it as many times as you like, it still doesn't make you right. Those of us who know what we're talking about know that Netburst was a disaster after Northwood, and should have been ended with the Northwood, which was the only successful time period of Netburst. Unless you think that continuing to make P4's, which were good competition for the K7's, while your competitor has the K8 makes good sense?:laugh: Would you not ridicule AMD for sticking with Skt. 939 for another 3 years, like Intel did with the P4, even though the C2D has been released?

Concerning spies, I think there probably are a few. I'll edit this post to include their names and titles, just as soon as I they are made publicly available.

It wasn't a disaster after Northwood. The Tech industry just shifted from Raw Power to efficiency. I challenge you to find another chip that scales from 1.6GHZ all the way up to 7GHZ (Not sure if it was 7, but was reached on Liquid N20).

It was a success before Prescott, when Prescott was released the market just went in a different direction.

If you want a failure, look NV3x architecture. They tried to do too much. Look at the Cyrix MII processors. Look at Slot Processor Design. I could go on and on...P4, more specifically Prescott, while not an engineering marvel was certainly not a failure.

-Kevin

 

[zsdersw]

Originally posted by: myocardia
Feel free to repeat it as many times as you like, it still doesn't make you right. Those of us who know what we're talking about know that Netburst was a disaster after Northwood, and should have been ended with the Northwood, which was the only successful time period of Netburst.

Here's a tip: if you're going to quote people when you respond, quote the exact part you're responding to.

The question about "disaster" with regards to Netburst after Northwood isn't something to be "in the know" about. It's a matter of perspective.

Was Netburst after Northwood a disaster in sales? Nope.. they sold just fine.

Was Netburst after Northwood a disaster in pure performance? Not really, unless you also consider C2D vs. K8 a disaster or nearly a disaster.

Was Netburst after Northwood a disaster in power usage/heat generation? Prescott/Smithfield could be argued as such, but Cedar Mill/Presler.. not really.. so a mixed bag with that one.

Was Netburst after Northwood a disaster on price/performance? Some offerings were, some weren't. The fastest of them wasn't worth the price, but the low-end units were.

Was Netburst after Northwood a disaster in engineering? Well, that brings up a whole other set of perspectives, but the most basic is no, it was not. Engineering disasters are those that don't work.. and/or fail prematurely. Netburst chips after Northwood worked just fine and had lifespans similar to any other CPU, regardless of what you thought of their performance or power usage/heat generation.

 

[CTho9305]

I'm sure they follow each other's patents and press releases.

As I understand it, trade secrets aren't patented immediately so that they don't have to be disclosed.

There is also probably a great deal of crossover among past employees at both companies.

There is definitely a lot of that. A lot of my coworkers are ex-Intel.

 

[SickBeast]

Originally posted by: CTho9305

I'm sure they follow each other's patents and press releases.

As I understand it, trade secrets aren't patented immediately so that they don't have to be disclosed.

There is also probably a great deal of crossover among past employees at both companies.

There is definitely a lot of that. A lot of my coworkers are ex-Intel.

What do you do at AMD? And what is the current sense there? Are you guys terrified about intel? Do you expect to take over the perfomance lead this year?