New Woodcrest Conroe tests at Anandtech

[BaronMatrix]

Woodcrest/T2000/Opteron


It shows the Woodcrest winning overwelmingly, though I do have a few questions, like why they mention the DL385 as he Opteron system but using an MSI system for the test bed.

Also, they seem to be runnign the benchmarks differently than the last time showing a generic number instead of concurrency and users resulting in transactions per second.

Also the systems only have 4GB of RAM now when I remember them having more. The DL 385 holds 16GB of 3200 DDR. It is looking like they are testing on a small server scale and not a large one.

I'm not doubting their veracity but it seems strange that all of a sudden the benchmark methods have changed without a formal notice. Can anyone be trusted when it comes to Opteron?

I know I expect to be flamed for beng a fanboy but I am a scientist so I like things to be ordered and level, using the same methodology or explaining why it changed. I do still hope that the numbers I saw in the last Paxville test improve, but they aren't doing the benchmarks the same. (I have been looking for the Paxville/Opteron article but it seems to have disappeared from AnandTech)

 

[Caboose-1]

In any CPU test between AMD and Intel, there will most always be indiscrepancies. The only way to get a correct test is to do it yourself, but that is most always out of the question because most folks don't have the cash to blow on two full systems just to do a test. Regardless of the results and indiscrepancies, I believe that the Woodcrest is "better," but the Opteron is still a good chip. I should know because a 185 Has been been serving me well in my high end system. Alas, this time arouind I will be buying an Intel. I hop the "fence," all the time, like a real PC builder should. There is no sense blindly supporting a company.

 

[BaronMatrix]

In any CPU test between AMD and Intel, there will most always be indiscrepancies. The only way to get a correct test is to do it yourself, but that is most always out of the question because most folks don't have the cash to blow on two full systems just to do a test. Regardless of the results and indiscrepancies, I believe that the Woodcrest is "better," but the Opteron is still a good chip. I should know because a 185 Has been been serving me well in my high end system. Alas, this time arouind I will be buying an Intel. I hop the "fence," all the time, like a real PC builder should. There is no sense blindly supporting a company.

I just like benchmarks to be "stable" or always the same. I will still recommend Opteron for 4Way, but I don't need a new server. Can you understand why they would intro one server as a test bed and then test with another? the beginning of the article mentioned the DL385, the tests all showed MSI.


I'm curious if anyone saw the last Apache tests between Paxville and Opteron on Anadtech? It seems to have disappeared.

 

[Caboose-1]

I just like benchmarks to be "stable" or always the same. I will still recommend Opteron for 4Way.....

With you on this .

 

[BaronMatrix]

I just like benchmarks to be "stable" or always the same. I will still recommend Opteron for 4Way.....

With you on this .

8)
I still hope to see a test with 16GB and actually using the DL385. A server with 4GB RAM is low end. Like I said I have 4GB in my 4400+.

 

[Panzerzero]

Sun OS's love ram and do a lot with it. I would think if they got a good AMD based server the test wouldn't look the same 8GB vs. 4GB on a Sun OS is a lot. This isn't even a test as a game. Based on the setup it was just play time with no real conclusions. Even in CHM 113 you need controls. Basic controls for any CPU test mem Size, hard drive(s), video, network card(when doing web test), internal cables(trust me), monitor tools, and case.

 

[9-inch]

Please check this thread out:
conrunt loses to X2 3800+

 

[shinigamiX]

9-Inch? INCH? You're still using INCHES? Puny human. I'm 1.3 LIGHT-YEARS. I could string you up like a spit roast, fool! You and every other retard on the face of this planet! And that's saying something.

Also, check this thread out:
The World hates 9-Inch, demands he STFU

 

[Action_Man]

Owned.

 

[Action_Man]

LOL, you got owned in the comment section! :lol: :lol: :lol: :lol:

 

[spud]

In any CPU test between AMD and Intel, there will most always be indiscrepancies. The only way to get a correct test is to do it yourself, but that is most always out of the question because most folks don't have the cash to blow on two full systems just to do a test. Regardless of the results and indiscrepancies, I believe that the Woodcrest is "better," but the Opteron is still a good chip. I should know because a 185 Has been been serving me well in my high end system. Alas, this time arouind I will be buying an Intel. I hop the "fence," all the time, like a real PC builder should. There is no sense blindly supporting a company.

I just like benchmarks to be "stable" or always the same. I will still recommend Opteron for 4Way, but I don't need a new server. Can you understand why they would intro one server as a test bed and then test with another? the beginning of the article mentioned the DL385, the tests all showed MSI.


I'm curious if anyone saw the last Apache tests between Paxville and Opteron on Anadtech? It seems to have disappeared.

Well sure they can stop changing the benchmarking techniques up when the hardware and software stop changing.

 

[Whizzard9992]

Sun OS's love ram and do a lot with it. I would think if they got a good AMD based server the test wouldn't look the same 8GB vs. 4GB on a Sun OS is a lot. This isn't even a test as a game. Based on the setup it was just play time with no real conclusions. Even in CHM 113 you need controls. Basic controls for any CPU test mem Size, hard drive(s), video, network card(when doing web test), internal cables(trust me), monitor tools, and case.

Not just sun, but the Java VM and RAM is like a fat kid and cake: you keep giving and it keeps taking Solaris has always been an impressive OS for Java apps (which is why they survive with such a high price point).

Even so, my position remains the same that Sun is too overpriced. I don't care how many cores they crunch onto a die. I'll take a blade server anyday.

 

[BaronMatrix]

In any CPU test between AMD and Intel, there will most always be indiscrepancies. The only way to get a correct test is to do it yourself, but that is most always out of the question because most folks don't have the cash to blow on two full systems just to do a test. Regardless of the results and indiscrepancies, I believe that the Woodcrest is "better," but the Opteron is still a good chip. I should know because a 185 Has been been serving me well in my high end system. Alas, this time arouind I will be buying an Intel. I hop the "fence," all the time, like a real PC builder should. There is no sense blindly supporting a company.

I just like benchmarks to be "stable" or always the same. I will still recommend Opteron for 4Way, but I don't need a new server. Can you understand why they would intro one server as a test bed and then test with another? the beginning of the article mentioned the DL385, the tests all showed MSI.


I'm curious if anyone saw the last Apache tests between Paxville and Opteron on Anadtech? It seems to have disappeared.

Well sure they can stop changing the benchmarking techniques up when the hardware and software stop changing.


Not the techniques, the UNITS. AND the test bed.

 

[spud]

In any CPU test between AMD and Intel, there will most always be indiscrepancies. The only way to get a correct test is to do it yourself, but that is most always out of the question because most folks don't have the cash to blow on two full systems just to do a test. Regardless of the results and indiscrepancies, I believe that the Woodcrest is "better," but the Opteron is still a good chip. I should know because a 185 Has been been serving me well in my high end system. Alas, this time arouind I will be buying an Intel. I hop the "fence," all the time, like a real PC builder should. There is no sense blindly supporting a company.

I just like benchmarks to be "stable" or always the same. I will still recommend Opteron for 4Way, but I don't need a new server. Can you understand why they would intro one server as a test bed and then test with another? the beginning of the article mentioned the DL385, the tests all showed MSI.


I'm curious if anyone saw the last Apache tests between Paxville and Opteron on Anadtech? It seems to have disappeared.

Well sure they can stop changing the benchmarking techniques up when the hardware and software stop changing.


Not the techniques, the UNITS. AND the test bed.

They included a Sun system in there what do you expect?

 

[BaronMatrix]

Owned.

Never happen. I'm not competing It would be unfair to you.

 

[Action_Man]

Right. Opteron and the AMD fanboys were owned.

 

[Whizzard9992]

ooooh action man changed his avatar... I almost didn't recognize ya! 8O

 

[BaronMatrix]

LOL, you got owned in the comment section! :lol: :lol: :lol: :lol:

I don't thinkso. I replied to his so-called "owning" that I was wondering why they didn't call up Intel for some more RAM. They got chips. Can you explain why they used MSI for test results but DL385 as a test bed?

 

[old_times]

I just like benchmarks to be "stable" or always the same. I will still recommend Opteron for 4Way.....

With you on this .

8)
I still hope to see a test with 16GB and actually using the DL385. A server with 4GB RAM is low end. Like I said I have 4GB in my 4400+.

Did you look at their software package? the 4 GB is way more than enough to load the entire package adding more memory won't help.


,,

 

[BaronMatrix]

I just like benchmarks to be "stable" or always the same. I will still recommend Opteron for 4Way.....

With you on this .

8)
I still hope to see a test with 16GB and actually using the DL385. A server with 4GB RAM is low end. Like I said I have 4GB in my 4400+.

Did you look at their software package? the 4 GB is way more than enough to load the entire package adding more memory won't help.


,,


yes I looked at the software package, but that has nothing to do with UNITS and less to do with why the DL385 had no results, unless it was a typo and their shouldn't be two MSI servers n the graph.

 

[ltcommander_data]

I have to leave soon, but I'd thought I'd give my take since I dropped by the forum.

like why they mention the DL385 as he Opteron system but using an MSI system for the test bed.

The reason for the switch from the DL385 to the MSI appears to be quite simple from my point of view. The DL385 is a four socket system loaded up with dual cores for a total of 8 threaded parallelism. It simply wouldn't be fair to compair a 2 socket setup to a 4 socket setup. The only reason why the added it in the initial benchmarks was to give a sense of scale. Anandtech was quite clear in this regard.

http://www.anandtech.com/IT/showdoc.aspx?i=2772&p=5

We included the HP DL585 to see whether 8 cores of complex general purpose CPUs (Opteron 880) can keep up with the 8 MAU of the Sun T1. If you want to compare Woodcrest and the Opteron, you should check the 2 and 4 concurrency numbers.

They included it just to show 8 threaded performance. They clearly indicate you should really be looking at the 2 and 4 threaded performance between Woodcrest and the 2 socket MSI setup.

Also the systems only have 4GB of RAM now when I remember them having more.

This is speculation on my part, but the reason why 4GB of RAM was used was because of the limited availability of FB-DIMMs. Especially, if Anandtech is pieceing their system together from various companies without official help from Intel (which is the independance we want of course). 4GB uses 1GB chips which are easier to obtain then say 2GB chips and gives 1 piece to each channel.

That said, it should be noted that only DDR2 533 memory was used in the FB-DIMM configuration. This leads to the ironic scenario where the FSBs are providing more bandwidth than the memory can produce. The concern is not the lack of bandwidth since the platforms are bandwidth constrained anyways as Anandtech noted, it's that DDR2 667 would offer lower latencies which is important in server applications. This would mean of course that there is still room for improvement.

 

[BaronMatrix]

I have to leave soon, but I'd thought I'd give my take since I dropped by the forum.

like why they mention the DL385 as he Opteron system but using an MSI system for the test bed.

The reason for the switch from the DL385 to the MSI appears to be quite simple from my point of view. The DL385 is a four socket system loaded up with dual cores for a total of 8 threaded parallelism. It simply wouldn't be fair to compair a 2 socket setup to a 4 socket setup. The only reason why the added it in the initial benchmarks was to give a sense of scale. Anandtech was quite clear in this regard.

http://www.anandtech.com/IT/showdoc.aspx?i=2772&p=5

We included the HP DL585 to see whether 8 cores of complex general purpose CPUs (Opteron 880) can keep up with the 8 MAU of the Sun T1. If you want to compare Woodcrest and the Opteron, you should check the 2 and 4 concurrency numbers.

They included it just to show 8 threaded performance. They clearly indicate you should really be looking at the 2 and 4 threaded performance between Woodcrest and the 2 socket MSI setup.

Also the systems only have 4GB of RAM now when I remember them having more.

This is speculation on my part, but the reason why 4GB of RAM was used was because of the limited availability of FB-DIMMs. Especially, if Anandtech is pieceing their system together from various companies without official help from Intel (which is the independance we want of course). 4GB uses 1GB chips which are easier to obtain then say 2GB chips and gives 1 piece to each channel.

That said, it should be noted that only DDR2 533 memory was used in the FB-DIMM configuration. This leads to the ironic scenario where the FSBs are providing more bandwidth than the memory can produce. The concern is not the lack of bandwidth since the platforms are bandwidth constrained anyways as Anandtech noted, it's that DDR2 667 would offer lower latencies which is important in server applications. This would mean of course that there is still room for improvement.


No, you're thinking of the DL585. That is the HP with 4 sockets. The DL 385 is a dual proc machine. In one or two tests they do have a 585 as a measure, but now 385.

 

[Parrot]

Woodcrest/T2000/Opteron


It shows the Woodcrest winning overwelmingly, though I do have a few questions, like why they mention the DL385 as he Opteron system but using an MSI system for the test bed.

Also, they seem to be runnign the benchmarks differently than the last time showing a generic number instead of concurrency and users resulting in transactions per second.

Also the systems only have 4GB of RAM now when I remember them having more. The DL 385 holds 16GB of 3200 DDR. It is looking like they are testing on a small server scale and not a large one.




I'm not doubting their veracity but it seems strange that all of a sudden the benchmark methods have changed without a formal notice. Can anyone be trusted when it comes to Opteron?

I know I expect to be flamed for beng a fanboy but I am a scientist so I like things to be ordered and level, using the same methodology or explaining why it changed. I do still hope that the numbers I saw in the last Paxville test improve, but they aren't doing the benchmarks the same. (I have been looking for the Paxville/Opteron article but it seems to have disappeared from AnandTech)

ROFLMAO!! Anandtech are a joke!! They have only succeeded in giving an object lesson in how NOT TO conduct server benchmarking-OR ANY BENCHMARKING!!!!!!
:lol: :lol: :lol: :lol::lol: :lol: :lol: :lol::lol: :lol: :lol: :lol:

 

[Caboose-1]

Woodcrest/T2000/Opteron


It shows the Woodcrest winning overwelmingly, though I do have a few questions, like why they mention the DL385 as he Opteron system but using an MSI system for the test bed.

Also, they seem to be runnign the benchmarks differently than the last time showing a generic number instead of concurrency and users resulting in transactions per second.

Also the systems only have 4GB of RAM now when I remember them having more. The DL 385 holds 16GB of 3200 DDR. It is looking like they are testing on a small server scale and not a large one.




I'm not doubting their veracity but it seems strange that all of a sudden the benchmark methods have changed without a formal notice. Can anyone be trusted when it comes to Opteron?

I know I expect to be flamed for beng a fanboy but I am a scientist so I like things to be ordered and level, using the same methodology or explaining why it changed. I do still hope that the numbers I saw in the last Paxville test improve, but they aren't doing the benchmarks the same. (I have been looking for the Paxville/Opteron article but it seems to have disappeared from AnandTech)

ROFLMAO!! Anandtech are a joke!! They have only succeeded in giving an object lesson in how NO TO conduct server benchmarking-OR ANY BENCHMARKING!!!!!!
:lol: :lol: :lol: :lol::lol: :lol: :lol: :lol::lol: :lol: :lol: :lol: What about all the other benchmarks where Woodcrest wins? Let me guess, they were ALL Intel floptimized, bribed, fake-ass benchmarks because it won out over an Opteron and Intel can't be trusted, they are crooked and rig everything![/shifty eyes]. Take off your blindfold.

 

[Parrot]

Woodcrest/T2000/Opteron


It shows the Woodcrest winning overwelmingly, though I do have a few questions, like why they mention the DL385 as he Opteron system but using an MSI system for the test bed.

Also, they seem to be runnign the benchmarks differently than the last time showing a generic number instead of concurrency and users resulting in transactions per second.

Also the systems only have 4GB of RAM now when I remember them having more. The DL 385 holds 16GB of 3200 DDR. It is looking like they are testing on a small server scale and not a large one.




I'm not doubting their veracity but it seems strange that all of a sudden the benchmark methods have changed without a formal notice. Can anyone be trusted when it comes to Opteron?

I know I expect to be flamed for beng a fanboy but I am a scientist so I like things to be ordered and level, using the same methodology or explaining why it changed. I do still hope that the numbers I saw in the last Paxville test improve, but they aren't doing the benchmarks the same. (I have been looking for the Paxville/Opteron article but it seems to have disappeared from AnandTech)

ROFLMAO!! Anandtech are a joke!! They have only succeeded in giving an object lesson in how NO TO conduct server benchmarking-OR ANY BENCHMARKING!!!!!!
:lol: :lol: :lol: :lol::lol: :lol: :lol: :lol::lol: :lol: :lol: :lol: What about all the other benchmarks where Woodcrest wins? Let me guess, they were ALL Intel floptimized, bribed, fake-ass benchmarks because it won out over an Opteron and Intel can't be trusted, they are crooked and rig everything![/shifty eyes]. Take off your blindfold.

Keep your shirt on!! I was commenting on their "methodology", although
"travesty" would be a better word. I have drawn no conclusions regarding
the performance of Woodcrest, let alone any other processor! For that, I shall wait for some PROFESSIONALS to do the work!!Believe it or not, I am interested in the figures but I need ACCURATE figures to work with,
NOT "toymarks" produced by some school kids!!! :lol: