HyperTransport and Opteron/Athlon64 Overclocking
The first question many will have about our efforts to look at how Athlon64 will perform is how we can possibly compare an overclocked Opteron to a chip that is not overclocked. In the case of the Opteron, the comparison is more accurate than you might first think.In normal setups (e.g. Athlon/P4), the CPU gets its clock from the FSB clock and multiplies it by the “clock multiplier” to determine how fast its internal clock should be. With a 16x multiplier, when the external clock ticks once, the CPU ticks 16 times. However, with the Athlon 64/Opteron, there is no FSB, so the CPU must get its clock from somewhere. It doesn't produce it internally; instead, it derives it from the native HT (HyperTransport) frequency, which is 200MHz, but because of the bus' nature, it runs at an effective 800MHz.
So, for our 1.8GHz Opteron 144, the multiplier is 9x, which is why raising the HT frequency to 222MHz increases the clock speed to around 2GHz. But we are increasing the HyperTransport clock in our overclocking, and not a FSB clock, which does not exist on Opteron/Athlon64. In real terms, this means our CPU overclocking has a significant impact on Performance, but it is unlikely that our increase in memory speed will have nearly as much impact on performance. Since we are nowhere near saturating the Hypertransport bus at 200 (effective 800), increasing HyperTransport to 222 (888) will not likely have much, if any, impact on overall performance. Our performance improvements, with Opteron/Athlon64, are mainly coming from increase in CPU clock — much more so than on the Pentium 4 or Athlon architectures.
Obviously, the PCI bus operates at a different frequency than the HT bus than the CPU, but they all operate based on multiples of each other, and are all derived from the HyperTransport clock.
Facebook
Twitter
RSS
79 Comments
View All Comments
Anonymous User - Sunday, September 7, 2003 - link
There are TPC benchmarks out with the new Opteron ship running 64-bit Linux and has gotten very good performance indeed.http://www.tpc.org/tpch/results/tpch_perf_results....
IBMs xSeries 325 is sporting an Opteron.
Wesley Fink - Sunday, September 7, 2003 - link
#66 -The Opteron and nForce3 require Registered ECC or Registered non-ECC dimms which are not the same as the unbuffered dimms we normally use for testing. It is very hard right now to find Registered dimms, let alone registered DDR400 dimms (the fastest ECC chips in current production are apparently designed for DDR333). The ONLY memory we had available that would work on the nForce3 were 512MB Registered ECC DDR400 modules. Dula-Channel makes that a megabyte. The theory goes that ECC and Registered slow down memory performance a bit compared to unbuffered, which should certainly offset any advantage.
Our benchmark tests are cumulative, as we certainly can't keep rigs up for every board we test. With a new setup like this and cross-platform data, we sometimes have to compare what we can get to what we already have.
Socket 754 will be single-channel and able to use regular unbuffered memory, while Socket 940 is dual-channel and requires Registered memory. The rumored Socket 939, scheduled for year-end, will allow use of standard unbuffered memory.
Locutus4657 - Saturday, September 6, 2003 - link
#27 Markets don't lie?? I remember when Rambuses stock was selling at $300/share and was rated BUY BUY BUY. If you're buying your tech prices based on stock values you are in some very seriouse trouble. And just so you know, I've built many an AMD system my self, and only the old "Super 7" systems had any real quirks. The newer Athlons have been problem free for me.Anonymous User - Saturday, September 6, 2003 - link
#58 - hey i hadn't noticed that before. wesley, what's up with 1gb on the Athlon64/Opteron adn only 512mb on the P4???Anonymous User - Saturday, September 6, 2003 - link
Some of you guys are so remarkably shortsighted it's not even funny. Posts like "haha, AMD still makes the crappier chip, they'll be out of business by this time next year" are foolish.If you like Intel, good for you, a strong AMD keeps intel's prices down and keeps the innovation coming. No AMD means we'll probably be paying $1200 for the 3.6 Ghz prescott this time next year, because they won't have to release as often or as cheaply.
Kudos to AMD, and I wish them nothing but the best.
Anonymous User - Saturday, September 6, 2003 - link
low voltage Itaniums may be lower in performance to their big Madison sisters, but they blow everything else out of the water on some industry standard SPEC scores!Anonymous User - Saturday, September 6, 2003 - link
Could it be that Registered DDR400 ECC Modules are as rare as the Athlon XP 2800+ that was paper launched at 2.25 GHz actual and then jumped quickly to a Barton 2800+ at 2.08 GHz? Or could it be that this is the same misleading stuff that AMD tried to do with their PriceWaterHouseCoopers benchmarks when they used the top of the line nForce platform against a mainstream Intel 845G platform and disabled Hyperthreading saying that it made the P4 slower - on top of that they used software patches that "aren't available for public download" to run their benchmarks on 3-year old benchmark software that was not optimized for the P4. If you look at what AMD has been saying for the past 5 years, they typically overpromise and underdeliver (a.k.a. Athlon FX - remarked Opteron now that Athlon64 will not be able to compete in it's original single-channel design) whereas Intel usually just stays pretty quiet, conservatively underpromises and overdelivers.Anonymous User - Saturday, September 6, 2003 - link
hey all you guys clamoring for linux benchmarks, speak up and mention some good ones for them to run...Anonymous User - Saturday, September 6, 2003 - link
I don't know why people are getting so emotional about intel vs amd. Let there be competition so we the consumers win. If Athlon xp hadn't been so powerful we wouldd not have 3.2 ghz, ht pentium 4 right now. On the other hand if we didn't have a 3.2 ghz p4 there would not be an upcoming athlon 64 fx. The chips with dual channel memory were all supposed to be in the server/workstation market. AMD has realised that the socket 754 chip will not be competitive enough with upcoming prescott and as such has created a new socket 939. I suspect that the socket 754 will go the way of the duron as a celeron competitor while socket 939 will be the flagship chip.Regarding the article the reviewer should have made it clear that this was an athlon 64-fx preview not athlon 64.
Anonymous User - Saturday, September 6, 2003 - link
The low voltage itaniums are also low performance.