Intel Woodcrest, AMD's Opteron and Sun's UltraSparc T1: Server CPU Shoot-out
by Johan De Gelas on June 7, 2006 12:00 PM EST- Posted in
- IT Computing
MySQL Results: Scaling
Back to our main subject, our astute readers have probably already noticed a weird anomaly. Let us analyze this further. If you look closely at both our measurements, Quad-core and Dual-core x86, you'll notice that the scaling is negative. To make it more clear, we made an average of all concurrency numbers from 5 and higher.
This is nothing short of amazing. It seems like an anomaly, but this is not the case. These benchmarks have been checked, verified and checked again. They are accurate. The x86 cores running on Linux perform better with two cores than with four cores, but the T1 running Solaris actually improves performance going from 4 to 8 cores.
So who is guilty? Linux or the Opteron system? We had to test with Solaris on the Opteron to be sure. However, the Serverworks chipset of our MSI 1U server was not supported by x86 Solaris. So we went back to our homebuilt server, based on the MSI K8N Master2-FAR.
And this puts the performance of our UltraSparc T1 in a whole different perspective. First of all, it is clear that while MySQL might not be the most scalable database, the current kernel of Linux is not helping matters. We did tweak the Linux kernel two ways: the 2.6.15 kernel was optimized for either Intel's or AMD's architecture and the AMD architecture also got NUMA support.
So what is going on here? After talking to our MySQL guru (P. Zaitsev), it turns out that in some circumstances, MySQL might cause trouble for the Linux mutex (mutual exclusion) implementation: "mutex ping-pong". The mutex implementation makes sure that two threads cannot access data in the main memory that is locked by another thread.
It seems however more a MySQL problem than a Linux one, as other databases like DB2 scale very well in Linux. For DB2 under the same load we noticed a performance increase of no less than 80-85% when going from two to four cores. Also, with some loads, the bad scaling kicks in later than our "Select dominated" load. Intel's performance labs told us that they also ran into the same problem.
These issues are not as severe as the problems we encountered with MySQL in Mac OSX. Note that Apple seems to have recognized the problem and seems to offer a workaround. We'll report back with other MySQL workloads to investigate the MySQL scaling problem further.
PostGreSQL Results
PostgreSQL 8.0.7, another open source database, uses processes and not threads to deal with connections. The consequence is that the benchmark numbers are a lot more stable: once each core is busy with it's process, you almost get maximum performance. In other words, the results didn't change much from 5, 10 or 25 concurrent users. To keep things simple, we only list the numbers with 20 users, which results in peak performance. The queries per second numbers at 5 and 25 were only a few percent lower. We did not include the T2000 Sun Server as the optimal PostGreSQL configuration is still under investigation.
Another clear victory for Woodcrest. On the Opteron, every 10% in clockspeed increase seems to result in a 7% performance increase. So if we extrapolate, an Opteron 3 GHz would arrive at 616 queries per second.
Back to our main subject, our astute readers have probably already noticed a weird anomaly. Let us analyze this further. If you look closely at both our measurements, Quad-core and Dual-core x86, you'll notice that the scaling is negative. To make it more clear, we made an average of all concurrency numbers from 5 and higher.
| MySQL Linux (Queries/s) | |||||
| Sun T1 4/8 cores 1 GHz |
MSI K2-102A2M Opteron 275 |
Xeon 5160 Woodcrest 3 GHz |
MSI K2-102A2M Opteron 280 |
||
| Average Dual-core (T1: quad-core) |
362 | 749 | 996 | 805 | |
| Average Quad-core (T1: octal-core) |
433 | 590 | 904 | 622 | |
| Speedup Dual to Quad | 20% | -21% | -9% | -23% | |
This is nothing short of amazing. It seems like an anomaly, but this is not the case. These benchmarks have been checked, verified and checked again. They are accurate. The x86 cores running on Linux perform better with two cores than with four cores, but the T1 running Solaris actually improves performance going from 4 to 8 cores.
So who is guilty? Linux or the Opteron system? We had to test with Solaris on the Opteron to be sure. However, the Serverworks chipset of our MSI 1U server was not supported by x86 Solaris. So we went back to our homebuilt server, based on the MSI K8N Master2-FAR.
| MySQL Solaris (Queries/s) | |||
| Sun T1 4/8 cores 1 GHz | Opteron 280 Solaris | Opteron 280 Linux | |
| Average Dual-core (T1: quad-core) |
362 | 456 | 799 |
| Average Quad-core (T1: octal-core) |
433 | 605 | 625 |
| Speedup Dual to Quad | 20% | 33% | -22% |
And this puts the performance of our UltraSparc T1 in a whole different perspective. First of all, it is clear that while MySQL might not be the most scalable database, the current kernel of Linux is not helping matters. We did tweak the Linux kernel two ways: the 2.6.15 kernel was optimized for either Intel's or AMD's architecture and the AMD architecture also got NUMA support.
So what is going on here? After talking to our MySQL guru (P. Zaitsev), it turns out that in some circumstances, MySQL might cause trouble for the Linux mutex (mutual exclusion) implementation: "mutex ping-pong". The mutex implementation makes sure that two threads cannot access data in the main memory that is locked by another thread.
It seems however more a MySQL problem than a Linux one, as other databases like DB2 scale very well in Linux. For DB2 under the same load we noticed a performance increase of no less than 80-85% when going from two to four cores. Also, with some loads, the bad scaling kicks in later than our "Select dominated" load. Intel's performance labs told us that they also ran into the same problem.
These issues are not as severe as the problems we encountered with MySQL in Mac OSX. Note that Apple seems to have recognized the problem and seems to offer a workaround. We'll report back with other MySQL workloads to investigate the MySQL scaling problem further.
PostGreSQL Results
PostgreSQL 8.0.7, another open source database, uses processes and not threads to deal with connections. The consequence is that the benchmark numbers are a lot more stable: once each core is busy with it's process, you almost get maximum performance. In other words, the results didn't change much from 5, 10 or 25 concurrent users. To keep things simple, we only list the numbers with 20 users, which results in peak performance. The queries per second numbers at 5 and 25 were only a few percent lower. We did not include the T2000 Sun Server as the optimal PostGreSQL configuration is still under investigation.
| PostgreSQL 8.0.7 (Queries/s) | |
| DL385 1 x Opteron 280 | 517 |
| Intel 2 x Xeon "Irwindale" 3.6 GHz | 448 |
| MSI 1U 1 x Opteron 275 | 490 |
| MSI 1U 1 x Opteron 280 | 524 |
| Intel 1 x Xeon 5160 WC 3 GHz | 673 |
Another clear victory for Woodcrest. On the Opteron, every 10% in clockspeed increase seems to result in a 7% performance increase. So if we extrapolate, an Opteron 3 GHz would arrive at 616 queries per second.
Facebook
Twitter
RSS
91 Comments
View All Comments
JohanAnandtech - Saturday, June 10, 2006 - link
The test you link is running apachebench while testing how fast STATIC html can be sent. Our LAMP test has to run PHP, access the MYSQL database, make calculations on that data ... this called DYNAMIC content.If you do not understand why a static HTML page can be served many times faster than a complex one with dynamic content, well...
You are basically saying that a test is wrong because it doesn't give the same results as another test which tests with different software, different dataset. Duh.
BasMSI - Wednesday, June 14, 2006 - link
I noticed Johan.But still, it's stupid to use and publish benchmark results from a test that can't handle/test the systems at their max.
Come on, get real, it's like testing a Lada and a Ferrari on a track that can't do more then 100KM/H and then state, look how well the Lada keeps up with the Ferrari.
Also, what's wrong with static HTML tests?
I see no harm in those, many websites are still static.
And you used them before to show how fast the Opterons where, so why not again?
Now we have absolutly nothing to compare or verify....so bogus test-results.
BrechtKets - Saturday, June 10, 2006 - link
Maybe you should check the author of the aces hardware article.
Also not that those tests were done with apachebench en the tests now have been done with httperf and and autobench...
FreakyD - Friday, June 9, 2006 - link
Dell has released some new servers with the new Intel Woodcrest platform. The pricing is less than for the older Netburst architecture servers... It looks like we'll have a price war on our hands, and of course AMD will end up losing that battle since Intel has lower production costs with higher volume.Also interesting to note, the 3.0Ghz Woodcrest Intel processor that was quite competitive in this review is the lowest end processor on the new Dell servers. Their highest end one is a 3.73 Ghz part. AMD's highest end dual core server processor is currently 2.6 Ghz. So there's additional performance gains for Intel vs AMD in a highest end server processor shootout.
I'm disappointed that AMD hasn't done more since they released the K8 architecture. AMD has also been slow to release their new server platform with Pacifica enhancements.
It's too bad that Dell has taken so long to begin using AMD in servers. They've held the performance lead for quite some time. With technology and market leaders changing so fast, they should have been faster to adjust their product lineup.
duploxxx - Friday, June 9, 2006 - link
duh my dear friend.... the dell servers you are pointing to can be checked where? link?you are mixing woodcrest that is at max 3000mhz and the dempsey 3.73 both on the same platform. dempsey is still no match for the woodcrest and opterons, so thats normal that the price tag is that low...... and its already dead before it is even launched
check this review, the dempsey is still wiped out on 90% of all the benches by an old architecture and certainly if you would check the power consumption/performance chart.
http://www.gamepc.com/labs/view_content.asp?id=xeo...">http://www.gamepc.com/labs/view_content.asp?id=xeo...
the proc cost of intel is certainly not lower than the amd ones... looking at the die size the woodcrest and conroe are bigger
@anand, those type of benches would be nice on a woodcrest, if you fail to give them now by "any reason" they will be available in the near future by other reviewers. so its always better to be the first :)
FreakyD - Friday, June 9, 2006 - link
Ahh, my mistake, thanks for the correction so nobody else gets the wrong idea. Once again I'm confused by Intel's naming and numbering scheme to not know exactly what's being sold.Aileur - Thursday, June 8, 2006 - link
This is a sad sad display. And i dont mean the review, i mean everybody bashing this article and each other like their lives depended on it.Its a cpu review on a hardware site, try to put it into perspective.
You read it, you draw your own conclusions if you want to, you go on with your life.
ashyanbhog - Friday, June 9, 2006 - link
Sure our lives dont depend on it,but Anandtech was a site you could rely onto get unbiased reviews. I have configured specs for atleast 25 machines based on Anandtech reports. Whenever somebody asked which CPU or someother part was better, I would suggest that they search for its review on Anandtech.
Even in the IDF conroe demo, Anandtech failed to identify some parts of the Intel setup that could have impacted performance, it was only after readers expressed their displeasure that Anandtech did a second review with the updates that should have actually been part of the Intel setup preview
If this new found low of Anandtech continues, I'll have to choose a different site to base my decisions on.
Also remember, Intel has previously used and continues to use Anandtech review of its processors in its analysts meet and at other places. As somebody pointed out, even a $0.15 swing in Intel share prices alters its valuation by one billion dollar!!! Intel could buy a handful of review reports by favoring advertising budgets for a fraction of that money.
Anandtech made my life a little easier by giving unbiased reviews, looks like I'll have to get back to comparing results from a few reviews as I used to do before I discovered Anandtech
Slappi - Thursday, June 8, 2006 - link
The Message is Clear...........Anand is getting paid by the big Intel.
Seriously.... you guys should at least TRY to hide your bias.
I mean months of setting up and you miss a known error that falsely reports extremely low dual OP. numbers?!?
Woodcrest ROCKS?~?~?
Something tells me that is gonna come back to bite you one day in the near future.
AnandThenMan - Thursday, June 8, 2006 - link
well ya gotta love this statement:That's two words LOL