Henri Richard explains why AMD failed to gain more marketshare

[Cryslayer80]

Exactly...

 

[fazers_on_stun]

Well that's why I couldn't figure out what i was to be looking at Cryslayer. That would be the last place on earth I would go for any scoop on Intel anything .

You misread his post - he was slamming Anandtech, not AMDZone ..

Of course, he probably doesn't know that Johan at AT is one of the more respected and knowledgeable reviewers out there, but just reflexively engages in slamming any site that mentions Intel favourably.

 

[Cryslayer80]

You misread his post - he was slamming Anandtech, not AMDZone ..

Of course, he probably doesn't know that Johan at AT is one of the more respected and knowledgeable reviewers out there, but just reflexively engages in slamming any site that mentions Intel favourably.

If you meant me, I was not ironic in that post, I really think that moderator is ignorant. Maybe it doesn't count for the whole AZone but...

 

[fazers_on_stun]

If you meant me, I was not ironic in that post, I really think that moderator is ignorant. Maybe it doesn't count for the whole AZone but...

Oh, OK - next time use quotes - helps avoid confusing what you're saying vs. somebody else 😛...

 

[roofus]

I just couldnt figure out what was linked that would be seen as credible/relevant. I could have missed it as I was distracted by the notion that in their minds anyone who isn't fawning over AMD is pro-Intel.

 

[Cryslayer80]

I just couldnt figure out what was linked that would be seen as credible/relevant. I could have missed it as I was distracted by the notion that in their minds anyone who isn't fawning over ADM is pro-Intel.

Abu Dhabi Men's College? :lol: JK

 

[roofus]

lol i didnt fix it fast enough. 🙁

 

[jennyh]

You missed what was credible about slamming the article? I read a lot of stuff that brought a lot of the article into question.

And it wasn't just the op, there were plenty of other posters picking up on points. It's just a typical example of intel bias, whether with wording or something else. As somebody pointed out...

Considering that AMD’s ACP numbers are rather optimistic and Intel’s TDPs are rather pessimistic, the 8-thread quadcore 1.86 GHz L3426 ($284) makes the six-core 1.8 GHz Opteron 2419EE look expensive ($989). The former can push it’s clock up to 3.2 GHz under single threaded loads, and is thus a really interesting option if your application has a significant part of non-parallel code.

And a comment from MU which is very valid..

5. The "competition offers better per core and per chip performance" comment also ignores the fact that Magny-Cours is coming out with far more cores per chip than anything Intel has and a 12-core Magny-Cours will offer better performance per chip than a 6-core Westmere in multithreaded server applications.

What is he basing all these ideas on? Nehalem-EX is huge, hot and hungry. It had better be extremely powerful. Better performance/watt than MC? I seriously doubt it in the majority of cases, but then again if you start out believing ACP is 'optimistic' and intels TDP is 'pessimistic' (seriously how much garbage is that? Intel downplaying their TDP? :lol: ), then sure you're gonna believe that Nehalem-EX is going to win on perf/watt. Kinda difficult to get over that initial bias, and the wording used in much of that caused a lot of consternation.

 

[jimmysmitty]

http://www.amdzone.com/phpbb3/viewtopic.php?f=52&t=137151&start=0&st=0&sk=t&sd=a

Yea.... no.

They don't even like Anand talking about Intel, AMDs main competition in the x86 market.

I love how back when AMD was on top and Intel was in the same article when talking about what AMD was coming out with and what Intel had to offer, no one complained. Now they all do.

You missed what was credible about slamming the article? I read a lot of stuff that brought a lot of the article into question.

And it wasn't just the op, there were plenty of other posters picking up on points. It's just a typical example of intel bias, whether with wording or something else. As somebody pointed out...

Considering that AMD’s ACP numbers are rather optimistic and Intel’s TDPs are rather pessimistic, the 8-thread quadcore 1.86 GHz L3426 ($284) makes the six-core 1.8 GHz Opteron 2419EE look expensive ($989). The former can push it’s clock up to 3.2 GHz under single threaded loads, and is thus a really interesting option if your application has a significant part of non-parallel code.

And a comment from MU which is very valid..

5. The "competition offers better per core and per chip performance" comment also ignores the fact that Magny-Cours is coming out with far more cores per chip than anything Intel has and a 12-core Magny-Cours will offer better performance per chip than a 6-core Westmere in multithreaded server applications.

What is he basing all these ideas on? Nehalem-EX is huge, hot and hungry. It had better be extremely powerful. Better performance/watt than MC? I seriously doubt it in the majority of cases, but then again if you start out believing ACP is 'optimistic' and intels TDP is 'pessimistic' (seriously how much garbage is that? Intel downplaying their TDP? :lol: ), then sure you're gonna believe that Nehalem-EX is going to win on perf/watt. Kinda difficult to get over that initial bias, and the wording used in much of that caused a lot of consternation.

No he clearly states Westmere 6 core. Nehalem-EX is no where to be mentioned. Biggest problem is you are basing everything on nothing. You have nothing to back up your claims about Nehalem EX.

ACP is just a load of crap really. They say a optimistic approach if you do normal loads. Well normally servers hit high loads a lot. I mean consitently. So how does ACP figure into that? If a server is under full load for 90% of the year, ACP is worthless.

Besides currently a single Xeon has no problems keeping up with a 2P Opty system so unless Istanbul increases IPC as well as core count Nehalem EX wont have that much trouble considering it will have 4x QPI and Quad DDR3 to enhance that memory loving server setup.

 

[jennyh]

ACP is just a load of crap really. They say a optimistic approach if you do normal loads. Well normally servers hit high loads a lot. I mean consitently. So how does ACP figure into that? If a server is under full load for 90% of the year, ACP is worthless.

And this johan at anand figured out all by himself that intel were 'pessimistic' over TDP? I can assure you all the intels pull a helluva lot more at max than any of the AMD's.

Besides currently a single Xeon has no problems keeping up with a 2P Opty system so unless Istanbul increases IPC as well as core count Nehalem EX wont have that much trouble considering it will have 4x QPI and Quad DDR3 to enhance that memory loving server setup.

It never fails to amaze me. You really do believe that every single thing intel releases is some kind of wonder miracle chip don't you? :lol:

Now I'm not saying Beckton is Larrabee, far from it. But it's quite close to Fermi I'll wager, and by that I mean too big, too hot and possibly too late. I wonder why we never see any Anandtech articles about that?

 

[fazers_on_stun]

You missed what was credible about slamming the article? I read a lot of stuff that brought a lot of the article into question.

And it wasn't just the op, there were plenty of other posters picking up on points. It's just a typical example of intel bias, whether with wording or something else. As somebody pointed out....

Yeah well I read it some months ago back when the controversy was fresh, and most of the comments amounted to style rather than substance, such as Johan's mentioning Intel too frequently in an AMD server roadmap review. Johan rebutted by stating that maybe the title was incorrect but that any realistic review of AMD's roadmaps would have to include comparison to its much bigger competitor - makes sense, no? Unfortunately the AMD fanbois including MU_Engineer came across as sharks seeking blood in the water, with "let's gang up on this idiot who dares invade our turf". Personally I lost a lot of respect for MU after reading his diatribe, which I think he did in order to up his creds with the idiots who run the forum. But that's just me - others can read your link and form their own opinions...

"Considering that AMD’s ACP numbers are rather optimistic and Intel’s TDPs are rather pessimistic, the 8-thread quadcore 1.86 GHz L3426 ($284) makes the six-core 1.8 GHz Opteron 2419EE look expensive ($989). The former can push it’s clock up to 3.2 GHz under single threaded loads, and is thus a really interesting option if your application has a significant part of non-parallel code."

And a comment from MU which is very valid..

"5. The "competition offers better per core and per chip performance" comment also ignores the fact that Magny-Cours is coming out with far more cores per chip than anything Intel has and a 12-core Magny-Cours will offer better performance per chip than a 6-core Westmere in multithreaded server applications."

What is he basing all these ideas on? Nehalem-EX is huge, hot and hungry. It had better be extremely powerful. Better performance/watt than MC? I seriously doubt it in the majority of cases, but then again if you start out believing ACP is 'optimistic' and intels TDP is 'pessimistic' (seriously how much garbage is that? Intel downplaying their TDP? :lol: ), then sure you're gonna believe that Nehalem-EX is going to win on perf/watt. Kinda difficult to get over that initial bias, and the wording used in much of that caused a lot of consternation

Unless MU works at Intel or AMD, which he doesn't, just how the heck does he know this on unreleased CPUs? He's just speculating like everyone else. Personally I'd put much more stock in some of the Intel engineers who post here and on Roborat's blog.

As for ACP vs. TDP, I'm sure you've seen the evidence on the other side by now, yet you choose to ignore same. AMD's 'ACP' is mostly marketing BS so that they can wedge themselves into a lower power budget.

The really sad part is that AMDZealotZone idiots manage to drive off any and all opposing viewpoints, no matter how respectfully offered, so there's no useful discussion there. Unless you count agreement between biased morons as 'useful' 😀

 

[fazers_on_stun]

And this johan at anand figured out all by himself that intel were 'pessimistic' over TDP? I can assure you all the intels pull a helluva lot more at max than any of the AMD's.

Based on your ownership of a Q6600, no doubt 😀. Pardon us for not taking your assurances on this, Jenny..

It never fails to amaze me. You really do believe that every single thing intel releases is some kind of wonder miracle chip don't you? :lol:

Now I'm not saying Beckton is Larrabee, far from it. But it's quite close to Fermi I'll wager, and by that I mean too big, too hot and possibly too late. I wonder why we never see any Anandtech articles about that.

Well lessee - you're not going to believe any favorable reviews of Intel products, when they do come out, so what's the point of arguing?? "All the benchmarks will have been rigged with Intel compilers" (even those that are shown to not be), or else "Intel bribed the review sites", yadda-yadda..

 

[BadTrip]

I don't think so Jenny.

Nehalem-EX will take care of Mangy Cours and Intel have already stated their intention of using 32nm Westmere to target the server market.

I am SOOOO EXTREMELY JEALOUS OF YOUR AVATAR!!!!! DAMN YOU!!! LOL

 

[jennyh]

Nah I think there were many valid points raised, but if you go into it with the attitude that it's going to be garbage then what do you expect to read?

MU is *easily* one of the most respected posters here, by a long shot btw. What he said was right - how the hell does this anand reviewer know that 'the competition offers better per core and per chip performance'?

Who the hell cares about 'per core' performance anyway...oh ye I get it, it's all those server guys running single threaded apps that make the nehalems look good isn't it?

If intel are so damn great where are their 12 core cpu's that make the most of many core tasks? There is a lot of wisening up to do from intel fans tbh, because MC is going to be cheap and awesome while you keep waiting on a hot, hungry and stupidly expensive chip that nobody really needs.

 

[Chad Boga]

Now I'm not saying Beckton is Larrabee, far from it. But it's quite close to Fermi I'll wager, and by that I mean too big, too hot and possibly too late. I wonder why we never see any Anandtech articles about that?

Because yet again things are not as you state and thus Anandtech don't want to appear moronic and without a clue.

 

[fazers_on_stun]

Nah I think there were many valid points raised, but if you go into it with the attitude that it's going to be garbage then what do you expect to read?

MU is *easily* one of the most respected posters here, by a long shot btw. What he said was right - how the hell does this anand reviewer know that 'the competition offers better per core and per chip performance'?

Who the hell cares about 'per core' performance anyway...oh ye I get it, it's all those server guys running single threaded apps that make the nehalems look good isn't it?

If intel are so damn great where are their 12 core cpu's that make the most of many core tasks? There is a lot of wisening up to do from intel fans tbh, because MC is going to be cheap and awesome while you keep waiting on a hot, hungry and stupidly expensive chip that nobody really needs.

First of all, there's a rather large difference between somebody who posts stuff here or AMDzone gratuitously, and an industry-respected reviewer who makes his living doing reviews. The professional reviewer has much more resources available to him or her, such as help from AMD or Intel when configuring or testing new products. You might notice that the actual AMD employee over at AMDZone, JF-AMD, is rather respectful of Johan, whereas the rest are not. Yes, there were probably some valid points raised about the tone of the article, but I didn't see much of anything substantial.

We can revisit this conversation in a year and see how much Intel gained in 4P+ marketshare, despite their 'hot, hungry and stupidly expensive chips'. 😀...

 

[jimmysmitty]

And this johan at anand figured out all by himself that intel were 'pessimistic' over TDP? I can assure you all the intels pull a helluva lot more at max than any of the AMD's.

Besides currently a single Xeon has no problems keeping up with a 2P Opty system so unless Istanbul increases IPC as well as core count Nehalem EX wont have that much trouble considering it will have 4x QPI and Quad DDR3 to enhance that memory loving server setup.

It never fails to amaze me. You really do believe that every single thing intel releases is some kind of wonder miracle chip don't you? :lol:

Now I'm not saying Beckton is Larrabee, far from it. But it's quite close to Fermi I'll wager, and by that I mean too big, too hot and possibly too late. I wonder why we never see any Anandtech articles about that?

We have always known that Intel likes to over shoot their TDP under full load. Its the way AMD did it before as well. Its not a bad thing to do when the TDP turns out to be lower by 30w or more than the rated TDP.

As for the "miracle" chip, you were the one going on about buying AMD would mean a cure for cancer.

Server apps, as we all know, loved K8 for its memory bandwidth. So when you take something that already has more memory bandwidth than K10, give it another channel and multiply the interconnect link by 4, what does that do? It gives a hell of a lot more memory bandwidth to play with and the server apps will eat it up and love it.

Unlike Istanbul which is just adding more cores or MC which is MCM (I still find it ironic that AMD laughed at Intel for the same thing yet here they are doing it........) of 2 Istanbul cores, Beckton adds more cache per core, increaded memory channels and much faster interconnects.

With all that I can say that I am pretty sure that a Beckton based server chip will have no problems going up against a MC chip. It wasn't to long ago that it would take near 2 Xeons in to compete with 1 Opty in certain situations, so why is it hard to believe that a 8 core CPU might be able to one up a 12 core CPU? Oh wait its Intel.......

As for its release, Intel never specified an exact date for its release. They said late 2009 or early 2010 and so far its still set for 3 months from now.

And remember, AMD was late with K10. By a lot. Being late doesn't mean anything. I do love your insider info though. You somehow know more about Beckton than anyone else, even review sites, and know that MC will be cheap. If its not.............

Meh.

 

[BadTrip]

All this late talk form the AMD camp needs to stop.

Dare I menton Bulldozer.

And dont start the "Intel cheated..." BS.

 

[JAYDEEJOHN]

If Intel doesnt get their heat together, theyll have a tough time taking new markets. Perf is nice, but heat n power are nicer. Theres something definately going on with Intel here, and looking at the 32nm at this point, with the voltage etc, makes me wonder if theyre having a tough time

 

[verndewd]

If Intel doesnt get their heat together, theyll have a tough time taking new markets. Perf is nice, but heat n power are nicer. Theres something definately going on with Intel here, and looking at the 32nm at this point, with the voltage etc, makes me wonder if theyre having a tough time

Intel chips are humanities only hope to avert nibiru and other extinction level events,,,,,,,,and bed bugs relief ,, yeah bed bugs. 😀

 

[JAYDEEJOHN]

Im just guessing here, as bumping V to 1.4 caused a flame out, and the rumors of heat issues etc.
But yes, we need those Intel chips for spongebobs square pants designs, the spotted owls perfect trees list etc.

 

[jennyh]

Server apps, as we all know, loved K8 for its memory bandwidth. So when you take something that already has more memory bandwidth than K10, give it another channel and multiply the interconnect link by 4, what does that do? It gives a hell of a lot more memory bandwidth to play with and the server apps will eat it up and love it.

Yet according to the anand article moving up to DDR3 for AMD will not give any real tangible benefits. Enough said? I think so.

Unlike Istanbul which is just adding more cores or MC which is MCM (I still find it ironic that AMD laughed at Intel for the same thing yet here they are doing it........) of 2 Istanbul cores, Beckton adds more cache per core, increaded memory channels and much faster interconnects.

What AMD said back then is irrelevant, both companies do plenty of bitching at each others methods. The important thing is we have this 'professional reviewer' making per-core comparisons when they simply aren't valid. Kinda funny that these comparisons are made now that AMD has more cores. Enough said?

With all that I can say that I am pretty sure that a Beckton based server chip will have no problems going up against a MC chip. It wasn't to long ago that it would take near 2 Xeons in to compete with 1 Opty in certain situations, so why is it hard to believe that a 8 core CPU might be able to one up a 12 core CPU? Oh wait its Intel.......

At no time did I say Beckton won't compete. They are clearly going to be very fast. However if you think you aren't going to pay silly amounts for them, think again.

As for its release, Intel never specified an exact date for its release. They said late 2009 or early 2010 and so far its still set for 3 months from now.

It always amazes me how intel get away with not setting exact dates for certain things. Is it any wonder they are never late in the eyes of the fanboys? :lol:

And remember, AMD was late with K10. By a lot. Being late doesn't mean anything. I do love your insider info though. You somehow know more about Beckton than anyone else, even review sites, and know that MC will be cheap. If its not.............

Meh.

You mean like with Larrabee? Oh wait we don't talk about that one any more do we. 😗

 

[jennyh]

If Intel doesnt get their heat together, theyll have a tough time taking new markets. Perf is nice, but heat n power are nicer. Theres something definately going on with Intel here, and looking at the 32nm at this point, with the voltage etc, makes me wonder if theyre having a tough time

I read something on semiconductor.net that basically suggested intel barely made any forward movement at 32nm compared to 45nm. You would expect it to get better over time but yeah, what should have given them a commanding lead hasn't materialised.

A weak, early 32nm is probably worse than intels very mature and strong 45nm. AMD are in no rush, they know the big increase in coming with HKMG.

 

[JAYDEEJOHN]

I waiting for all the praise to come out with 8 cores etc. To minimize this is ridiculous, and could be attributed to cppu envy

 

[Reynod]

henri worked on the ipad design ... not the marketing campaign ... from the looks of it from here anyway.