I´m abit late for commenting on these comments but here I go.
First of all, it is very sad to see these ongoing accusation towards Anandtech about beeing Intel-lover etc.
Personally, for me, I see the article as a whole, not just one page or one chart with numbers.
I read up, think and decide if the article has a good value in terms of things mentioned, considered etc.
I personally am an Intel fan. Even during the days with the Pentium 4. Though, to my defense I was abit immature at the time. Didn´t "want" AMD to win. But that´s another story.
I can now clearly recommend AMD based products if they meet the criteria for the purchase to be done.
For the servermarket it´s abit more difficult.
You have to look at much more numbers, manufacturer, ownership cost, spare parts, support things etc...
Johan - as always, I appreciate your expertise on these things, particularly in the area of virtualization and recognizing some of the subtle differences in the two architectures.
* Why is there no mention regarding the cost of DDR3 memory? It's important to call out that customers are going to have to pay a premium for it over DDR2 for at least the remainder of 2009.
* What about power consumption at the wall? That's quickly becoming the number one buying criteria these days.
* The 45 percent advantage with SMT? Where are you getting this number? Is it from a benchmark you ran or did Intel provide you with this?
" Why is there no mention regarding the cost of DDR3 memory? It's important to call out that customers are going to have to pay a premium for it over DDR2 for at least the remainder of 2009. "
I currently have no idea what buffered DDR-3 is going to cost, but looking at the unbuffered stuff, it is doubtfull if DDR-3 is really going to have a large impact on your server price.
Power Consumption: we'll measure this in detail, but why would you assume that there will a large difference? Previous server comparisons have shown that most of the difference came from large amount of FB-DIMM and Intel won't use them on the Nehalem EP.
"
* The 45 percent advantage with SMT? Where are you getting this number? "
I wrote "Up to 45 percent". But cases of 25-30% higher performance are pretty common. That is huge. Just imagine what kind of clockspeed or IPC advantage you must have to counter this.
Not a large impact? Good unbuffered DDR3 costs over twice as much as good unbuffered DDR2.
Buffered DDR3 will be somewhat a niche produkt, while memory companys starving to death, you expect it to be cheap? This is finally a product they can make money with and you bet they will do.
Power Consumption no difference? Seen all these desktop reviews where Nehalem leads the pack from the negative standpoint?
Conclusion: everything you know, you know from Intel. Otherwise you would tell us basic stuff like power consumption at the wall.
Your blog posts have the value of Intel press releases and should be considered as those.
I believe everything in the server market has a premium price. because the lack of information, all we can do about the memory price is to wait.
About the power consuption, you are only partially correct. Yes, Nehalem is a power hog, let's assume it consumes 30% more power than Core2 and Shangai, but it also do the work 50% faster (rough numbers). So make the simple calculation below.
previus CPU's: 100w over 5minutes; nehalem: 130W over 2.5 minutes. Job done, CPU goes to idle. Idle power is the same for both platforms. wich one will cost more money by the end of the month???
You are right, Mr Finally. I thank you for correcting me. I did a mental calculation and got trapped in the "2.5 is 50% of 5", but for sure the right math was hard enough that you couldn't do it yourself. It is easy to blame others and add nothing in the comments here. I wonder why do we have such passionate posts anyway.
But let's focus on the technology instead of blaming on each other. If anyone else wants to know the "math", here is the "bug-fixed" example:
Others CPUs: 100w over 5minutes, Nehalem: 130W over 3,3334 minutes. So total power is... 5x100 = 500 and 3.3334x130 = 433,3.
Ops, I guess Nehalem did it again. 66,7W cheaper than any other CPU when doing the same work. Watts per work done should be the measure, not peak power.
lol I just have to say finally sounds like an amd fanatics. oh well what else is new :P
/joke
seriously even you count the idle power consumption its hard to overcome the more than 10% difference in performance/watt estimated from the "bug-fix" calculation since both procs idles at a fraction of peak power usage.
and on DDR3, I guess what the author means is that it will not have a big impact on "server" price. DDR3 is a lot more expensive than DDR2 when used in a desktop application. but servers, especially those for HPC application are a lot more expensive to begin with. so the marginal increase might not be too big. also given that DDR3 is really on free fall in terms of pricing right now it's likely they'll be a lot lower towards later part of the year
ultimately I think AMD has great value in certain segment of the market. but I think we all have to tip the hat to intel as it owns the performance crown at the moment. this is from a person still using amd on his own rig :P
And I must say I'm a fan of good products. It took Intel over 5 years to come with a CPU (core2) that was better than the AMD K8 (Athlon 64), we all should thank AMD for firing up this competition - because we, consumers, receive the benefit from this.
By the way, the GPU on my rig is made by AMD. In fact, ALL the GPU's I bought in the last 4 years were AMD. =D
This article had good information but was written poorly. The person who wrote this should probably go to back and take a few English classes; sorry if that sounds harsh, but come on, this is anadtech.com. I expect better.
This isn't an article, it's a blog post. But this is exactly why articles and blog posts should be *clearly separated* on AnandTech! Articles are held to a different (higher) standard, and mixing your blog posts in with your articles is lowering the overall reputation of this venerable site.
Pleading "guilty with extenuating circumstances" :-). I promised I would follow up the previous blogpost within a day. So I finish it with a jetlag while waiting for my next plane to arrive. Bad plan. I hope I got most of the irritating errors. If not, mail me, I'll appreciate it.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
Facebook
Twitter
RSS
17 Comments
Back to Article
Rackmountsales - Tuesday, February 24, 2009 - link
Thanks for sharing information and i appreciate it.Looking for more discussion and waiting for new topics here.Rackmountsales
mikaels - Monday, February 23, 2009 - link
I´m abit late for commenting on these comments but here I go.First of all, it is very sad to see these ongoing accusation towards Anandtech about beeing Intel-lover etc.
Personally, for me, I see the article as a whole, not just one page or one chart with numbers.
I read up, think and decide if the article has a good value in terms of things mentioned, considered etc.
I personally am an Intel fan. Even during the days with the Pentium 4. Though, to my defense I was abit immature at the time. Didn´t "want" AMD to win. But that´s another story.
I can now clearly recommend AMD based products if they meet the criteria for the purchase to be done.
For the servermarket it´s abit more difficult.
You have to look at much more numbers, manufacturer, ownership cost, spare parts, support things etc...
AMDOpteronPhil - Friday, February 13, 2009 - link
Johan - as always, I appreciate your expertise on these things, particularly in the area of virtualization and recognizing some of the subtle differences in the two architectures.* Why is there no mention regarding the cost of DDR3 memory? It's important to call out that customers are going to have to pay a premium for it over DDR2 for at least the remainder of 2009.
* What about power consumption at the wall? That's quickly becoming the number one buying criteria these days.
* The 45 percent advantage with SMT? Where are you getting this number? Is it from a benchmark you ran or did Intel provide you with this?
JohanAnandtech - Friday, February 13, 2009 - link
" Why is there no mention regarding the cost of DDR3 memory? It's important to call out that customers are going to have to pay a premium for it over DDR2 for at least the remainder of 2009. "I currently have no idea what buffered DDR-3 is going to cost, but looking at the unbuffered stuff, it is doubtfull if DDR-3 is really going to have a large impact on your server price.
Power Consumption: we'll measure this in detail, but why would you assume that there will a large difference? Previous server comparisons have shown that most of the difference came from large amount of FB-DIMM and Intel won't use them on the Nehalem EP.
"
* The 45 percent advantage with SMT? Where are you getting this number? "
I wrote "Up to 45 percent". But cases of 25-30% higher performance are pretty common. That is huge. Just imagine what kind of clockspeed or IPC advantage you must have to counter this.
balancedthinking - Monday, February 16, 2009 - link
Not a large impact? Good unbuffered DDR3 costs over twice as much as good unbuffered DDR2.Buffered DDR3 will be somewhat a niche produkt, while memory companys starving to death, you expect it to be cheap? This is finally a product they can make money with and you bet they will do.
Power Consumption no difference? Seen all these desktop reviews where Nehalem leads the pack from the negative standpoint?
Conclusion: everything you know, you know from Intel. Otherwise you would tell us basic stuff like power consumption at the wall.
Your blog posts have the value of Intel press releases and should be considered as those.
marc1000 - Tuesday, February 17, 2009 - link
I believe everything in the server market has a premium price. because the lack of information, all we can do about the memory price is to wait.About the power consuption, you are only partially correct. Yes, Nehalem is a power hog, let's assume it consumes 30% more power than Core2 and Shangai, but it also do the work 50% faster (rough numbers). So make the simple calculation below.
previus CPU's: 100w over 5minutes; nehalem: 130W over 2.5 minutes. Job done, CPU goes to idle. Idle power is the same for both platforms. wich one will cost more money by the end of the month???
Finally - Wednesday, February 18, 2009 - link
50% faster is not 100% faster. You just halved the time.Learn to do the math or cease bullshitting.
marc1000 - Wednesday, February 18, 2009 - link
You are right, Mr Finally. I thank you for correcting me. I did a mental calculation and got trapped in the "2.5 is 50% of 5", but for sure the right math was hard enough that you couldn't do it yourself. It is easy to blame others and add nothing in the comments here. I wonder why do we have such passionate posts anyway.But let's focus on the technology instead of blaming on each other. If anyone else wants to know the "math", here is the "bug-fixed" example:
Others CPUs: 100w over 5minutes, Nehalem: 130W over 3,3334 minutes. So total power is... 5x100 = 500 and 3.3334x130 = 433,3.
Ops, I guess Nehalem did it again. 66,7W cheaper than any other CPU when doing the same work. Watts per work done should be the measure, not peak power.
Finally - Wednesday, February 18, 2009 - link
Hehe. Guess that I actually DID the math to make my comment.Once again you fail @ suggesting things.
In your faulty "bugfixed" calculation you seem to think that a Nehalem system in idle mode uses ZER0 power. Yeah, sure...
So, what's next?
Will Nehalem prove to be a perpetuum mobile?
Really. This whole what-if is written by an Infel fanatic for Intel fanatics... but what's the news?
gomakeit - Wednesday, February 18, 2009 - link
lol I just have to say finally sounds like an amd fanatics. oh well what else is new :P/joke
seriously even you count the idle power consumption its hard to overcome the more than 10% difference in performance/watt estimated from the "bug-fix" calculation since both procs idles at a fraction of peak power usage.
and on DDR3, I guess what the author means is that it will not have a big impact on "server" price. DDR3 is a lot more expensive than DDR2 when used in a desktop application. but servers, especially those for HPC application are a lot more expensive to begin with. so the marginal increase might not be too big. also given that DDR3 is really on free fall in terms of pricing right now it's likely they'll be a lot lower towards later part of the year
ultimately I think AMD has great value in certain segment of the market. but I think we all have to tip the hat to intel as it owns the performance crown at the moment. this is from a person still using amd on his own rig :P
marc1000 - Wednesday, February 18, 2009 - link
I could not agree more, gomakeit, perfect words./cheers
And I must say I'm a fan of good products. It took Intel over 5 years to come with a CPU (core2) that was better than the AMD K8 (Athlon 64), we all should thank AMD for firing up this competition - because we, consumers, receive the benefit from this.
By the way, the GPU on my rig is made by AMD. In fact, ALL the GPU's I bought in the last 4 years were AMD. =D
befair - Thursday, February 12, 2009 - link
Johan praising Intel ... old news .. next?Johnmcl7 - Thursday, February 12, 2009 - link
Seems the cat is out of the bag...Johnmcl7 - Thursday, February 12, 2009 - link
"The AMD chip is still disadvantaged by the fact that it does have SMT."Hrel - Thursday, February 12, 2009 - link
This article had good information but was written poorly. The person who wrote this should probably go to back and take a few English classes; sorry if that sounds harsh, but come on, this is anadtech.com. I expect better.LokutusofBorg - Thursday, February 12, 2009 - link
This isn't an article, it's a blog post. But this is exactly why articles and blog posts should be *clearly separated* on AnandTech! Articles are held to a different (higher) standard, and mixing your blog posts in with your articles is lowering the overall reputation of this venerable site.JohanAnandtech - Friday, February 13, 2009 - link
Good suggestion.Pleading "guilty with extenuating circumstances" :-). I promised I would follow up the previous blogpost within a day. So I finish it with a jetlag while waiting for my next plane to arrive. Bad plan. I hope I got most of the irritating errors. If not, mail me, I'll appreciate it.