Busting AnandTech's Woodcrest vs Opteron benchmark
Anand reproduced MySQL bug #15815, and quickly termed it a feature called "negative scaling"
We have proven that Anand is a paid Intel pumper here. Now, AnandTech has done a series of Woodcrest vs Opteron benchmarks. Again, AnandTech showed Intel having substantial lead.
Really? Can a paid pumper be trusted ever again? Let's forget about the past, and look at AnandTech's newest presentations without prejudice.
One year ago, in June 2005, AandTech's Johan De Gelas did similar benchmarks on dual core Opteron. You can see from that page that Opteron MySQL performance was 17% lower under Gentoo Linux than under SuSe Linux. On the same page, AnandTech compared the MySQL (InnoDB) performance of two (2) dual core Opteron 875 processors against one (1) dual core Opteron processor under Gentoo Linux. The conclusion was that adding a processor led to about 10% total performance increase. This was extremely poor scaling, but there was at least some increase.
AnandTech's system configuration on Woodcrest and Opteron is here. Pay attention to the motherboard AnandTech chose for the Opteron. It was the K8N Master2-FAR. What's special about this board? It is probably the only 2P Opteron board with only one bank of memory. With this board, the second Opteron always has to get its memory from the first CPU, incurring high latency. This MSI board (pictured below) is definitely not designed for server use, it doesn't even have onboard video.
A typical Opteron 2P server board looks likt this Tyan S2892, with two banks of memory, one for each CPU. The only AMD recommened MSI Opteron board is the K8D Master3-FA4R.
For both Woodcrest and Opteron, Gentoo Linux 2.6.15-gentoo-r7 and Mysql5.0.21 were used. Gentoo has about 1% of the Linux market and is getting more popular. Now, look at the MySQL performance numbers AnandTech got here.
First, we look at the comparison between one(1) 3GHZ Woodcrest Xeon 5160 and one (1) 2.4GHZ Opteron 280. The Woodcrest did 996 queries/second, while the single Opteron did 805 queries/second. The Woodcrest had 25% clockspeed advantage and 23% lead over the Opteron 280 on MySQL query speed.
Now, look at the next row, which was TWO Woodcrest CPUs against TWO Opteron 280 CPUs. The result? The 2P Opteron 280 system showed a 23% performance DECREASE from 1P Opteron 280. It did 622 queries/second. Adding a CPU in a 2P system leads to 23% drop in total MySQL performance? Two processors slower than one?
Anyone with half a brain would immediately conclude that something was seriously wrong with this benchmark. The result was simply stupid. If some moron told his boss that adding a CPU actually causes performance drop, he should be fired immediately. Because he should have used his brain to fixed the sucker instead of making such stupid conclusions. You need to dig deeper, you need to debug the software and see where it's hanging, and then solve it. MySQL has a lot of parameters to tune, so does the Linux SMP kernel. You either try out all available combinations to find a rational outcome or you use your brain to identify and fix the problem. Of course, doing this requires deeper understanding of the operating system and MySQL. We don't expect AnandTech's people possess this level of knowledge, otherwise they'd be working as software or system engineers who make a lot more money and accumulate a lot of stock options. But, they should at least contact the OS and DB vendor and Google the WEB* to find a solution. If the problem was identified as a major bug in the OS or DB or was otherwise persistent, they should try some other OS such as RedHat Enterprise, SuSe Enterprise, or different versions of OS and DB. A nonsense software result simply can't be used to represent CPU performance. A nonsense result only makes the benchmarker look stupid.
Yet, AnandTech concluded that "[q]uad-core and Dual-core x86, you'll notice that the scaling is negative... It seems like an anomaly, but this is not the case. These benchmarks have been checked, verified and checked again. "
So, not only AnandTech is a paid pumper, they are also surprisingly incompetent*.
Stupidity lends no confidence.
Actually, if you look at the first graph on this page, you can see that the 2P Opteron 280 quickly reached a saturation point at concurrency level of 2. But the 1P Opteron 280 scaled up quite nicely up to concurrency level 8. Apparently, there was a resoucre contention problem in the 2P setting. You just need to dig deeper...
Performing a Google search with the words "Linux opteron dual core MySQL bug", I quickly found this MySQL.com's page for bug #15815. The #15815 bug is highly reproducible, it was a bug due to mutex locks and cache line "ping-pong". There was a fix for it. In the discussion of the "negative scaling" on AnandTech, "mutex ping-pong" was mentioned. This is indication that someone at AnandTech actually knew about this. At the bottom of the bug #15815 page, a user reported:
"I'm using MySQL 5.0.21 on a 2.6.16 Gentoo system, I've tried both Gentoo compiled and precompiled Mysql distribution, and the Xeon machine (dual xeon 3.8Ghz with 8GB RAM) acts too much better then the dual Opteron (Opteron 275 with 16GB RAM)...better as 640 tps (Xeon) versus 110 (!!!) tps for the Opteron."
Note this guy was using the Gentoo OS and the same MySQL 5.0.21* Anand was using. He was also running dual Opteron 275.
Maybe someone noticed this bug report on mysql.com and instructed Anand to use the same setup to reproduce it but present it as an Opteron benchmark?
* The most current version of MySQL is 5.0.22.
posted by Sharikou, Ph. D. @ 6/07/2006 04:53:00 PM
50 Comments:
Oh My Sweet Dear God....that is almost as stupid as this (Link).
You know, I did notice a slow down on my server when I added a 2nd Opteron processor....oh wait, it doubled in speed, sorry.
/emote Anand is a moroff
Wow, Sharikou, you have reached new lows. Adding a new processor results in a 23% drop in performance relative to the first. Even with HTT, things won't scale perfectly, and that 23% represents the "non-perfect" scaling. It doesn't mean overall box performance goes down, it just means that if you have two Opterons, 1x = 100% performance, and an extra one adds 77% more performance relative to the first.
Seriously. Some of the stuff you write is downright frightening, but at least you pretend to have some facts. This incompetency is beyond anything I have ever seen before, though.
To the poster below my first one, you sir, are a bigger moron than Anand. The test is not showing scaling performance comparisons in the sense you're talking about, it is showing the numbers of 4-Cores vs. 2-Cores, and it shows that 4-Cores performs below 2-Cores, which is just f*ckin' stupid.
Adding a new processor results in a 23% drop in performance relative to the first. Even with HTT, things won't scale perfectly
Dude, use your brain. If the situation was as you desribed, that the second core has added 77% performance, why did AnandTech tardies call that an anomaly? How do you explain the Solaris 10 entry with 20% increase? You think that would indicate 120% performance increase by adding a CPU?
Go back and read the original carefully, the data was on this page:
http://www.anandtech.com/IT/showdoc.aspx?i=2772&p=8
"two Opterons, 1x = 100% performance, and an extra one adds 77% more performance relative to the first."
I think right math should be:
2 X 77% = 154%. so this dual opteron system gets 54% performance boost with this application instead of 100% in fantasy.
This benchmark looks very fishy from the ground up. Although there's a lot of information, it's presented in such a way to distract your attention from the real facts.
First of all, this tries to pose as a legitimate Linux benchmark but it only specifies that a 64bit kernel was used. There are no details on the Kernel configuration. The Linux Kernel has a few parameters that can be tweaked in order to favour a certain architecture.
I am somewhat disappointed that Sharikou tried to dismiss Gentoo only because it's only used in about 1% of all distributions. I have been using it for almost two years and I wouldn't use any other distro for anything in the world. Gentoo comes in stages. If you build a system from the first stage, you get to configure the compiler to use optimisations for a specific architecture. That means that if you build a system with the flags for the K8 on 64bits, you would have every binary on that system optimised for K8.
There were some debates on how big this difference is (and although I can't find the links right now) this accounted for about 10%.
Anand should have stayed with 64bit Windows XP and those nice colorful bars. Just because I see the words "64bit Linux" and some graphs generated straight out of Excel to make it look more professional doesn't impress me.
I would rather wait until the products are available for independent benchmarks and they all show the same facts before making up my mind. All this is speculation and it seems that the Intel propaganda machine is doing it's job quite well.
I am somewhat disappointed that Sharikou tried to dismiss Gentoo only because it's only used in about 1% of all distributions. I have been using it for almost two years and I wouldn't use any other distro for anything in the world. Gentoo comes in stages.
Sorry for that. I am not familiar with Gentoo. But AnandTech's previous result gave SuSe a better light than Gentoo. Of course, we know all Linux kernels are pretty much the same, though their default configurations may differ.
ACTUALLY, THE LINK "comparison between one(1) 3GHZ Woodcrest Xeon 5160 and one (1) 2.4GHZ Opteron 280" ON YOUR SITE IS REFLECTING A GRAPH OF "comparison between one(2) 3GHZ Woodcrest Xeon 5160 and one (1) 2.4GHZ Opteron 280" THIS IS CLEARLY INDICATED ON THEIR 2NG GRAPH OF THE PAGE. YOU CAN FOLLOW THIS LINK "http://www.anandtech.com/IT/showdoc.aspx?i=2772&p=8" TO DOUBLE CHECK. I AM NOT SURE IF THE GRAPH IS TYPO ON ANAND'S PART OR THE GUY IS TRYING TO UNFAIRLY BEAT DOWN AMD, I WOULD ONLY SUGGEST THAT ANAND SHOULD CONDUCT THE SAME TEST AGAIN USING REDHAT LINUX (WHICH IS MORE STANDARD).
All we need is someone from AMD to verify or dispute all these "independent testers".
And it doesn't help when AMD is so secretive about the performance of their 65nm parts.
Send out some engineering samples, even if it's an earlier build, so one can get some ideas.
somehow anand is.. being very generous with the benchmarks. i spoke with one of the intel representatives in computex yesterday, and he said anand basically increased the numbers on the benchmarks without notifying intel. that is one of the reasons why anand was called back to redo the benchmarks.
even though anand may pumped intel's benchmark numbers, i have to say intel won this round. i saw what conroe can do with my own eyes, after i made sure both Conroe and FX systems are competitive.
in other words, there is no way for FX/x2 to catch up with Conroe, and i believe its the case with Opteron and Woodcrest. AMD better come up with something better than 4x4, or they'll lose the market again to intel.
LOOK AT THIS GRAPH FROM ANAND, IT IS CLEARLY SAYING THEY ARE COMPARING (2 )WOODCREST 3.0GHZ AGAINST (1) OPTERON 275 AND 280.
"http://images.anandtech.com/reviews/it/2006/woodcrest-linux/MySQL5dual_woodcrest.gif"
THIS IS NOT A FAIR TEST!
Anandtech lost all credibility in my eyes long ago. He is as much of a sellout as Tom's Hardware.
Anand cleverly twisting some benchmarks to favor Intel for a few bucks paid under the table and the "exclusive" scoop on these products (brings lots of readers and ad revenue to his site) does not surprise me in the least.
actually, I would make the argument that evaluating only one version of Linux in this type of situation is not a good idea in and of itself. Not to knock Gentoo directly, it is a fine distro to itself, but Sharikou is correct in that it has a very small slice of the Linux market. However, it would have made more sense for Anandtech to have benchmarked using other distrobution types for a couple of reasons.
The first reason is the ability to duplicate the tests. This is actually a strike against Gentoo for what the operating system is. While it possible to duplicate an installation of Gentoo and the applications used, generating an exact copy of the exact configuration used without clear description of the compile targets used is very hard. This means that anybody wishing to reproduce these results on their own will be very hard-pressed to do so.
The second reason is commercial and residential use. Gentoo has it's market, that market just isn't very widespread. It would have made more sense for Anandtech to have tested a RPM based distro such as Mandriva, RedHat, Fedora Core, Novell Suse, or OpenSuse against a .deb based distro such as Debian(sid), Ubuntu, Mepis, or Xandros. The reason why it would have made more sense is that .deb and .rpm distros are actually used in the commercial and residential spheres, and used in great quantities. Had Anandtech used a distrobution that is in active use it would mean more to buyers currently looking to replace their Windows computers with a new system.
It would only be in the interests in providing a point of perspective that one would test a different type of Linux distrobution like Gentoo or Slackware.
Going back to the first point, had Anandtech benchmarked these on a Debian based system it would be fairly easy to duplicate the tests. Anandtech would just need to list the base version of the Debian distro they used, list the apt-repositories they pulled from, and the application in apt that were pulled. Anybody else who comes along afterwords with a Debian based distro would easily be able to duplicate the steps and the benchmarks.
The overall point is that while it is nice to see a non-dedicated Linux site approaching hardware, this isn't the way to approach it. As it stands now, the Anandtech tests are useless, reguardless of whatever results the benchmarks returned.
Wonder why AnandTech is using Microstar motherboards for the Opteron systems.
I thought there are better ones like Asus or Tyan for the Opteron platform.
The configurations for the test systems are not too detailed.
Mad Mod Mike, I don't think that your link is a 4x4 product...
But I really hope that some mobo company attempts at creating an Intel 4x4 equivalent.
I can just picture benchmarks comparing the two products...
AMD's 4x4 : 1,600 TPS
Intel's 4x4 : 600 TPS
Lets all pray for this product!
From some of the comments, I found a lot of people could not follow what AnandTech was doing and saying. What Anand found was this: with a 1P Opteron 280, you get 805 queries/s, with 2P you get only 622 queries/s for the whole system. Anand's idiots even give it a name "negative scaling".
As I found out (see footnotes), this phenomenon was most probably a bug or a configuration issue in Gentoo default+MySQL5.0.21, as other people are having the same issue.
Why use the MSI K8T motherboard for the Opterons? That have to be the worst motherboard they could have used for the opterons. It has a single memory controller; both processors do not have independent memory slots. This could cause a huge hit in performance.
Gentoo is a source based distribution. Which means the binaries will be compiled from source during the time of installation.
At the bare bare minimum, the kernel, the bootloader, and a few apps must be compiled. If you want to stray from the "defaults", the compiler switches to optimize for performance can be daunting.
Gentoo is not for everyone. It gives you a lot of ropes to play with, including the ability to hang yourself... this is not your father's distribution...
Me?... I'm leaving Fedora for Unbuntu.
think right math should be:
2 X 77% = 154%. so this dual opteron system gets 54% performance boost with this application instead of 100% in fantasy.
It's a waste of time to respond to this, but if I don't, some people will think the above interpretation of Anand's result was right.
Let me make it clear: AnandTech first had 1 Opteron 280, then they added another one. Instead of seeing some improvement, the total system performance dropped to 77% of the 1P level. The idiots called this "negative scaling". They probaly thought they could publish a paper of this "anomaly" and get a community college diploma.
Stupid benchmarks like this won't cause me to uninstall my HP AMD blade system and VMware servers for some woody thing..
All IT departments wait for 1-2 years of anyones new server launch, they wait for the fools to rush in and get burnt and then listen to the grape vine.. this is How AMD has displaced Intel from server rooms all over the world...Its going to years before Intel will displace all the AMD investment, unless they come up with a HyperTransport compatible connection they are HISTORY - end of story.
I think Anand deliberately used the hard-founded software package to complicate re-testing his results by others.
I also noticed that Intel server has two GB NIC interface while Opteron only one. How servers was connected to a switch is very important because this is a client-server congiguration.
If follow further for Anand, 8-ways server should be worse than 4, and 16 worse than 8.
So the best choice is a one-way server.
Ha-ha-ha.
Following the negative scaling of Anand, I think a clusters of Opteron would be as fast as an 486. LOL Anand.
Anand should have dug deeper on the drop in performance and not just stated what he saw.
Having said that no evidence has been brought up that Anand is paid by Intel to pumop the products. Some sort of financial evidence is needed rather than just a hjyperlink to where you say it before but still provide no evidence.
This sort of unsubstantiated remark has led to people being sued before now so you need to provide evidence I'm afraid or else it just looks like you are shooting the messenger.
Well, be fair to ANANDTECH. Finding the worst combination stuffs for AMD benchmark is not easy. ...Guys, do not under-estimate anandtech's technical knowledge.
Ha ha, check out mysql bug #15815 I googled out. They must have found it too.
This sort of unsubstantiated remark has led to people being sued before now so you need to provide evidence I'm afraid or else it just looks like you are shooting the messenger.
Look, Intel folks are visiting this Journal everyday, and I also bet Anand himself has read it. Intel has an army of lawyers. Am I afraid of being sued into bankruptcy?
No! First, I won't need a lawyer, I will fight myself till the US Supreme Court if needed, and I will enjoy doing it. Secondly, Intel or Anand are not stupid. They know what the truth is. Their hope is that fewer people will know the truth, and they will be as quiet as possible.
I have proven that
1) Anand is an Intel pumper
2) Anand has been paid by Intel directly and/or indirectly.
Now, I would further claim that Anand's increasing level of pumping is correlated to Intel's increasing level of paying. For this, I will have to do wide range of discovery in a potential lawsuit. I will have to get the depose various Intel and AnandTech people, I will have to request production of various invoices, payments over some period....
Anand should have dug deeper on the drop in performance and not just stated what he saw.
Anand cleverly twisting some benchmarks to favor Intel for a few bucks ...
I don't know why everyone is bashing Anand for this?
The guy is brilliant! He wanted to make a few bucks and has managed to do so -- at AMD's expense.
He didn't have to disclaim, google or dig deeper because that wasn't the purpose of his exercise.
Now having said that... Sharikou did his part in finding AND reporting about this Anand scam.
But more energy should be spent explaining the error/flaw and HOW it can be fixed so that those looking for a 2P system does NOT buy an Intel one because [s]he is too scared to confront this problem. (There are alot of LAZY techies out there).
At least that's my 2 cents...
Leave poor Anand alone. He's busy building some super eco-unfriendly McMansion, and must have run into some snags along the way. So Intel bails him out and he publishes one B.S. article, we would all do it for the amount of money Intel must have paid. I wouldn't build the McMansion, but I'd take the cash. Who really cares what his site says? We will all have the chance to do our own testing in the not so distant future.
Adding to DBA's summary of ANAND's frauds.
Anand mentioned that during the test he switched the cache of MySQL off. I think it emphasizes the impornace of large cache of Woodcrest and speed of IO controller. In real situation every admin would increase a cache size => Woodcrest's cachesize would become negligible regarding to a workload (see Sharikoa's comments about Conroe performance earlier), but lower latencies of Opteron would become a big advantage.
I love how you lot all lost the plot here. You're more interested in what Anand did or didn't do rather than the heart of it all which is that the new Conroe core from Intel is a good piece of architecture. I know you AMD lovers/fanboys out there just want the smallest victory you can, whether that's just because you don't want to believe it or have some sick tie to AMD but face it, you've had your time in the limelight, now it's Intel's turn to wipe those smug grins you've had on your face since Prescott's release. Fact was, if Core Duo wasn't enough to make a change Core 2 will do. All this crappy blog has done is try to find holes in Intel's technology and architecture and it's now world-reknowned across forums as being a joke and probably written by someone who works for AMD's spin machine.
the specint2000 results for the xeon 5160 (NGMA) is suspicious. they score over 3000, yet not even itanium2 does that. in addition opteron 256 (3ghz, single core scores over 2000) but all dual core opterons score less than that (all less than or equal to 2.6ghz) so i snooped aroung and found this article at grid-it.cnaf.infn.it/fileadmin/Certification/Spec.pdf which states that specint2000 is only supposed to be done on a single cpu, and if the cpu has dual core, the other core is NOT supposed to be used, i.e. turn it off in the configs. so i looked at other cpu results with multiple cpus and/or multicore like http://www.spec.org/osg/cpu2000/results/res2006q1/cpu2000-20060220-05633.html
and they ALL indicate that the settings were for a singel cpu and/oir core. i.e. they turned off all cores save one. BUT that is not indicated for the Dell systems using NGMA xeon! why don't they indicate this????????? because they didn't turn those off!!!!!!!!
You're more interested in what Anand did or didn't do rather than the heart of it all which is that the new Conroe core from Intel is a good piece of architecture.
The truth of Anand is important, because Anand's words were used by Intel as proof that Conroe is good. see
The Anand Story.
I have analysed the CORE2 architecture elsewhere in this blog. Essentially, CORE2 is 4 generations behind K8. All Intel tried to do is the same old thinking, pushing single core performance. AMD can defeat CORE2 in sleep: 4x4 is just like a slap of finger, and instantly Conroe is at 50% of AMD's performance. This is architectural superiority of Direct Connect.
Yes it is 4 generations behind k8 yet amd sees the need to copy the architectural advantages of the core2 in their next k8l release. I guess going 4 generations backwards is a new trend in the processor industry.
Yes it is 4 generations behind k8 yet amd sees the need to copy the architectural advantages of the core2 in their next k8l release.
Dude, it's the other way around. K8L has been in design for a very long time. Henri Richard said Intel's CORE2 copied AMD. It's just like Intel copied AMD's x86_64 after four Intel teams failed to make 64 bit work on x86.
But there are a lot of stuff INTEL hasn't figured out how to copy. look at K8L, it's so much advanced than anything out there. Intel stuff really seem primitive..
I guess every other hardware review sites that favor Core 2 Duos over AMD CPUs are paid Intel pumpers.
1: Sharikou, you're a moron! You completly turn the article your way.
2: The MySQL tests were done on Gentoo AND SUSE.
3: You didn't know about the MySQL bugreport until it was posted on anandtech.
4: scaling was negative on the woodcrest as wel.
5: they didn't write off opterons, it is said in the article that sometimes the opteron keeps up with the woodcrest (a much newer cpu), while sometimes the opteron is crushed (which is normal als the woodcrest is NEW).
6: there were other tests done than MySQL tests.
do you want me to go on??
scaling was negative on the woodcrest as wel.
This only shows AnandTech kids are even dumber. Two architecures, CORE2 and K8, both running MySQL slower in 2P than 1P. CORE2's drop is less. It's quite apparent this is a MySQL bug related to cache contention.
Yet, Anand touted as his discovery of "negative scaling".
"I have analysed the CORE2 architecture elsewhere in this blog. Essentially, CORE2 is 4 generations behind K8. All Intel tried to do is the same old thinking, pushing single core performance. AMD can defeat CORE2 in sleep: 4x4 is just like a slap of finger, and instantly Conroe is at 50% of AMD's performance. This is architectural superiority of Direct Connect."
And AMD64 (K8) is just a K7 on steriods as well?
Core 2 is the combination of the best of the "P6" attributes and the "Netburst" architecture. It introduces some major new additions such as sincle cycle SSE and other improvements over any current CPU out there (desktop market).
4X4 is accesible to how many peoples wallets? You need to have a special motherboard and then have to have 2X FX series (AMD has stated only FX will work) plus the necassary amount of RAM allocation to get it working, this is not to mention the huge increase in power consumption. And I am so sure if AMD was able to defeat Core so easily then why are they dropping their prices like crazy? Perhaps they don't believe they chips until Revision G will be able to keep up? Most of K8-Ls additions are great for the server oriented market, but to completely take away the lead of Core (Kentsfield by then) is beyond me.
[q]Yet, Anand touted as his discovery of "negative scaling".[/q]
No, Anand touted this as: Do not buy a quad core machine when you want to make use off MySQL. They did not suggest that this was the case for all software, and did not even suggest that it was the case for all database servers. If you had read the comments on the last article (about the t2000), you would have read that benches with DB2, Sybase and MS SQLSever are on the way because these compagnies prommised support on tuning the databases. The also said Oracle didn't provide support yet.
Why is it so important that the databases are completely tuned?? Because if not, your test would be completely useless. If your database is not tuned, you will not take advantage of your machine's capabilities.
The fact that MySQL scales negative, is a factor that people should keep in mind when they buy a quad server. There were also other tests done.
I think we should also wait and read the next article. I doubt the other databases will have similar results (scaling).
About the test methods you and other people complained about. This test (that you could not confirm) has been used in previous articles, and I didn't hear anyone complain when the Opterons performed a lot better than the old Xeons. But now that the woodcrest puts up resistance, the tests must be fraud! Can you explain that to me?? seems a bit like opportunisme, when the tests go your way their OK, if they don't they can't be real.
My 2 cents.
No, Anand touted this as: Do not buy a quad core machine when you want to make use off MySQL.
Well, Anand concluded that 2P Woodcrest is 45% faster than 2P Opteron 280. He wrote "As our extrapolated calculations show, even a 2.6 GHz Woodcrest will outperform the current Opteron 285 with a 5 to 55% margin, nothing short of impressive."
The question is, will big enterprises believe such low IQ, semi-retarded crackpots?
In this case a bug is a bug, not the "negative scaling" feature Anand used to pump Woodcrest. A fix for the bug has been available for quite a while now.
leaving the MySQL aside, there were also other test were the woodcrest outperformed the Opterons! Also tests where the opteron comes close.
I don't understand why you would make such a fuss over this. Isn't it normal that a that much older opteron would perform worse than the woodcrest? If a new opteron comes available, they will probably test it to, so if you say it's not fair to test an old opteron over the woodcrest, we will be able to say the same when a new opteron comes out.
To be fair I think the results weren't that good on anands tests for the woodcrest as other articles hype
actually, there is no patch for this problem! it will be solved in MySQL 5.1.
There is a workaround in mysql 5: set the innodb_thread_concurrency lower than 4. If they did that in the article it would result in worse performance!
FYI:
You got a reply from the author of the Anandtech article, Author: Johan De Gelas here:
http://aceshardware.com/forums/read_post.jsp?id=120058129&forumid=1
You got a reply from the author of the Anandtech article, Author: Johan De Gelas here
A bunch of cheap excuses for AnandTech's incompetence. The MySQL guy wrote "As you can see MySQL could be slower on Linux with many connections with increasing number of CPUs. This is known bug which being worked on." This was categorically different from AnandTech's brain dead conclusion. Software has bugs, that's the way of life, big deal. Instead of immediately recognizing the behaviour as a bug, AnandTech repeated the same thing multiple times, and then concluded that it was not anomaly but a case of "negative scaling". I didn't do this kind of test before, however, the minute I saw the AnandTech's report, I knew it must be a bug, and I googled it out in one minute. Anyone with a brain would do that.
Looking at AnandTech's response at
http://aceshardware.com/forums/read_post.jsp?id=120058129&forumid=1
you have to conclude that they also have very bad memory. Regarding their reply "Wrong. Dualcore Opteron was compared with two single core Opterons." to my summary of their previous test. Read this page again, scroll to the bottom of the page, in the last table, there were comparisons between "Dual dual core 875" and "single dual core 875", there is a column " Dual Dual core vs One Dual core" which shows about 10% scaling from adding a CPU.
Sharikou,
Unfortunately negative scaling does exist for MySQL with Innodb storage engines and there is no general workaround known to this date.
It could be using Gentoo was poor choise and there could be some other errors but this one is unlikely error.
If you get hold of Woodcrest yourself let me know and we can validate their findings :)
Gentoo is a verry tweekable distro, and they could deoptimize gentoo for the opteron!! and there was a bug in gentoo's version of mysql that redused preformance
Sharikou is not Ph.D., but a troll. He knows nothing actually but pretend he knows something. He is a charlaton . All his benchmarks are the lies.
A person who claims he got a Ph.D., but does not know adding a core can increase performance, such basic knowledge, much be a charlaton. Sharikou, you'd better told us where you got your Ph.D, or we will consider you troll.
If you claim you are a Ph.D., why don't you tell us which university you get it from?
Post a Comment
<< Home