Irate Anandtech readers pacified

WE REPORTED last week that an Anandtech review - Linux and EM64T; Intel's 64-bit Suggestion - hadn't been very well received by many who had commented on the article. The forums of Ace's Hardware and Slashdot were quick to highlight its shortcomings. With all that negative publicity raging, Anandtech listened to its readership and worked to make amends. Three days after the original piece was published, a new article - Linux [64-bit] Shootout: Opteron 150 vs. Xeon 3.6 Nocona - was released for public consumption.

Contrast two paragraphs from the original and follow up pieces:

"Without a doubt, the 3.6GHz Xeon trounces over the Athlon 64 in math-intensive benchmarks. Intel came ahead in every severe benchmark that we could throw at it, particularly during John the Ripper. Even though John uses several different optimizations to generate hashes, in every case, the Athlon chip found itself at least 40% behind. Much of this is likely attributed to the additional math tweaking in the Prescott family core." (1).

"After all is said and done it became difficult (nearly impossible?) to justify the Xeon processor in a UP [uni-processor] configuration over the Opteron 150, but perhaps we will see significant changes in dual and four way configurations." (2).

If you didn't already know, the Opteron 150 beat the 3.6GHz Xeon Nocona in 13 out of 17 tests. Now that pretty much turned the tables. Since the author's later conclusions were understandably a little less assertive than before, we'll address that imbalance here.

Winning benchmark margin for Xeon:
Chess
1. TSCP -03 -march - 2.6%
Synthetic
2. Opstone Scalar Product
Mean 32-bit - 7.9%
Mean 64-bit - 12.6%
Peak 32-bit - 17.1%
Peak 64-bit - 23.3%
Compression
3. GZip - 0.007%
Encryption
4. John the Ripper Blowfish
Config 1 - 32.0%
Config 2 - 13.8%
Config 3 - 14.8%

Winning benchmark margin for Opteron:
Database
1. MySQL Select - 22.0%
2. MySQL Insert - 21.3%
3. PostgreSQL Select - 40.7%
4. PostgreSQL Insert - 39.9%
Chess
5. TSCP -02 - 9.1%
6. TSCP -03 - 2.3%
7. Crafty - 69.7%
Rendering
8. mental ray - 16.2%
9. POV-Ray - 39.5%
Synthetic
10. Opstone Sparse Scalar Product
Mean 32-bit - 17.4%
Mean 64-bit - 13.6%
Peak 32-bit - 19.6%
Peak 64-bit - 19.8%
Encoding
11. Lame - 8.9%
Encryption
12. John the Ripper Des
Config 1 - 44.4%
Config 2 - 58.7%
Config 3 - 59.3%
13. John the Ripper MD5
Config 1 - 21.4%
Config 2 - 9.3%
Config 3 - 8.2%

Anandtech did run the OpenSSL benchmark but didn't provide any analysis - it did provide the raw data results. (3). AMD does have SPECweb99_SSL results for 2P servers which shows that Opteron just screams. (4).

As the author noted: We only used a 32-bit benchmark during the synthetic analysis, but still on SuSE 9.1 Pro (x86-64).

Based on these benchmarks, Opteron is once again in a different league. Intel's latest round of platform improvements still hasn't delivered. How long can the chip giant continue to trot out the line that says: "I think the theme is the overall platform elements"…"the complete platform"…"The message you should walk away with is that this really isn't about one technology or another, it's about delivering a collection of them".

What good does having "the complete platform" when the engine that drives it is second class? If Opteron could run on the Netburst architecture, Intel could rightly extol the virtues of its supporting technologies. But that is never going to happen. If the chip giant wants to compete in the premier league, it's going to have to come up with the goods.

Kudos to Anandtech for doing the right thing. If only others - Microsoft - would follow that example. µ

See Also
Irate AnandTech readers turn on reviewer
Microsoft gaming PC page still fails to deliver

L'INQ
New AnandTech piece
Pacified comments from article's forum
1
2
3
4