According to the EE times, AMD is coming out with Hyper Link 3.0 around June. Hyper Link 3.0 is supposed to have 2x the bandwidth of Hyper Link 2.0. This could be responsible for some gains that AMD expect out of AM2, without improvements in the actual processor (as Anandtech has reported).
HyperTransport is not used for memory access in a single socket configuration, in multiple socket configurations, memory requests can travel across HyperTrasport links.
2 days and it will be a full month since his last contact with us...
Anand, I know you are busy and all, but you are the driving force of this site. Without you, it will die...
Where in the world is... Carm- Anand Lal Shimpi?
Whose fault is it? CPU itself? Integrated Memory Controller? Or Motherboard implementation? It's somewhat relieving - I just invested heavily on S939 platform. ;) - but at the same time worrying, of course.
Will we see many enthusiasts moving to conroe/merom platforms at the end of the year?
Integrated memory controller, and maybe the fact that typical DDR2 667 memory has a higher bandwidth but a longer delay (latency) time until data can be fetched.
Tests (by Anandtech and others) have shown that Athlon64 is fairing better with lower latencies and less bandwitdh than the reverse. The performance "gap" between Socket 754 and Socket 939 (where you can see a doubling of bandwidth for a 10% some improvement) is proof enough.
DDR2 could prove usefull in multitasking/multicore scenarios
I thought it would be the reverse, since with more apps, less of both apps can fit in the cache effectively. Low latency access to the main memory would thus be necessary to make up the gap.
I still remember an article interviewing some AMD guy, saying that they would skip DDR2 and move directly to DDR3.
In my testing of various ram configs and others testing I've found A64 loves low latency and barley benefits from bandwidth..thus I've been guessing for over a year DDR2 won't do anything for AMD and for people to buy LL DDR setups now when the "should I wait for AM2" comes up. I would'nt expect any gain at all running sloppy timings and unless running DDR2-667 @ 3-2-2 1T it will not beat DDR1-400 @ 2-2-2 1T ...
forgetaboutit if you get BH5/UTT and clock DDR-500+ keeping same timings. Sanda pumps out around 8200 there and latency is about 36ns. DDR2 has no chance.
The only benefit AMD will see is lower system power consumption which is good with conroe coming. The only benefit consumer will see is lower prices since DDR is starting to escalate due to phase out, less heat, power.
Anand, what about the comments by Charlie over at the Inq that the current stepping of the AM2 platform have a major memory issue? Do you have the latest steppings?
The INQ is claiming that OEM sources say that the Rev F / AM2 combination will offer more than TEN percent (10%) clock/clock performance improvement in desktop / gaming.
"The parts are out there, and we are hearing various things, all centering around about a 10% performance gain, clock for clock. Server folks with 1207 parts tell us the gain is lower, desktop and gaming folks are aiming higher. It could just be variants among pre-release parts, or it could be the memory RAS taking a bite out of latency on the server parts. Either way, look for a bump."
Given that most software is still single threaded, the two cores aren't likely starved with dual DDR400. Benchmarks with the unsupported DDR500 dividers further prove that point.
Did Anand measure the memory bandwidth as halved compared to S939?
I can see how performance might be unchanged (or even inferior, as was the case with Prescott), but how on earth can the memory *bandwidth* go down?
Or should I just have paid more attention when I read the article? :)
"Needless to say my excitement vanished after I ran the first performance numbers and it offered about half the memory bandwidth of an average Socket-939 platform."
Yeah I haven't been expecting much from the AM-2 move. The way I see it, the move to AM-2 is more a result of DDR prices than performance. It won't be long that similar performing DDR2 will be cheaper than DDR1, so AMD wants to enable it's users to choose the more cost-effective platform. AMD CPU's have never been very memory bandwidth-hungry, so I didn't expect a speed increase. As Anand says, in the future, when we have quad core CPUs and are running several intense background apps/programs at the same time, then bandwidth becomes an issue. As we stand today, we *generally* use somewhere between 1 and 3 threads at a time, and AMD's memory controller simply doesn't need the huge bandwidth that high speed memory offers.
We’ve updated our terms. By continuing to use the site and/or by logging into your account, you agree to the Site’s updated Terms of Use and Privacy Policy.
19 Comments
Back to Article
kleinwl - Thursday, March 30, 2006 - link
According to the EE times, AMD is coming out with Hyper Link 3.0 around June. Hyper Link 3.0 is supposed to have 2x the bandwidth of Hyper Link 2.0. This could be responsible for some gains that AMD expect out of AM2, without improvements in the actual processor (as Anandtech has reported).Questar - Friday, March 31, 2006 - link
I assume you mean Hypertransport?I'm going to also assume that you know that HT has nothing to do with memory?
PaulDriver - Sunday, April 9, 2006 - link
bzzzzt.HyperTransport is not used for memory access in a single socket configuration, in multiple socket configurations, memory requests can travel across HyperTrasport links.
microAmp - Monday, March 27, 2006 - link
Any chance for some benchmarks on TES:O?creathir - Monday, March 27, 2006 - link
Anand is once again MIA... pushing 3 weeks now...Is this just...
"noanandtech.com"???
Hope all is well...
- Creathir
Nimbo - Thursday, April 6, 2006 - link
Officially one month without trace of Anand in the site that carries his name. I feel like an orphan.Houdani - Wednesday, April 12, 2006 - link
Not true. Anand's been posting articles, just not sharing about the quirks of his new house which should probably be done by now.creathir - Tuesday, April 4, 2006 - link
2 days and it will be a full month since his last contact with us...Anand, I know you are busy and all, but you are the driving force of this site. Without you, it will die...
Where in the world is... Carm- Anand Lal Shimpi?
- Creathir
lopri - Tuesday, March 7, 2006 - link
Whose fault is it? CPU itself? Integrated Memory Controller? Or Motherboard implementation? It's somewhat relieving - I just invested heavily on S939 platform. ;) - but at the same time worrying, of course.Will we see many enthusiasts moving to conroe/merom platforms at the end of the year?
Calin - Thursday, March 23, 2006 - link
Integrated memory controller, and maybe the fact that typical DDR2 667 memory has a higher bandwidth but a longer delay (latency) time until data can be fetched.Tests (by Anandtech and others) have shown that Athlon64 is fairing better with lower latencies and less bandwitdh than the reverse. The performance "gap" between Socket 754 and Socket 939 (where you can see a doubling of bandwidth for a 10% some improvement) is proof enough.
DDR2 could prove usefull in multitasking/multicore scenarios
Zoomer - Sunday, April 9, 2006 - link
I thought it would be the reverse, since with more apps, less of both apps can fit in the cache effectively. Low latency access to the main memory would thus be necessary to make up the gap.I still remember an article interviewing some AMD guy, saying that they would skip DDR2 and move directly to DDR3.
Zebo - Monday, March 6, 2006 - link
In my testing of various ram configs and others testing I've found A64 loves low latency and barley benefits from bandwidth..thus I've been guessing for over a year DDR2 won't do anything for AMD and for people to buy LL DDR setups now when the "should I wait for AM2" comes up. I would'nt expect any gain at all running sloppy timings and unless running DDR2-667 @ 3-2-2 1T it will not beat DDR1-400 @ 2-2-2 1T ...forgetaboutit if you get BH5/UTT and clock DDR-500+ keeping same timings. Sanda pumps out around 8200 there and latency is about 36ns. DDR2 has no chance.
The only benefit AMD will see is lower system power consumption which is good with conroe coming. The only benefit consumer will see is lower prices since DDR is starting to escalate due to phase out, less heat, power.
redpriest_ - Monday, March 6, 2006 - link
Anand, what about the comments by Charlie over at the Inq that the current stepping of the AM2 platform have a major memory issue? Do you have the latest steppings?dougSF30 - Monday, March 6, 2006 - link
The INQ is claiming that OEM sources say that the Rev F / AM2 combination will offer more than TEN percent (10%) clock/clock performance improvement in desktop / gaming.http://www.theinquirer.net/?article=30042">http://www.theinquirer.net/?article=30042
"The parts are out there, and we are hearing various things, all centering around about a 10% performance gain, clock for clock. Server folks with 1207 parts tell us the gain is lower, desktop and gaming folks are aiming higher. It could just be variants among pre-release parts, or it could be the memory RAS taking a bite out of latency on the server parts. Either way, look for a bump."
JackPack - Monday, March 6, 2006 - link
We didn't even see 10% from S754->S939.Given that most software is still single threaded, the two cores aren't likely starved with dual DDR400. Benchmarks with the unsupported DDR500 dividers further prove that point.
Spoonbender - Monday, March 6, 2006 - link
Did Anand measure the memory bandwidth as halved compared to S939?I can see how performance might be unchanged (or even inferior, as was the case with Prescott), but how on earth can the memory *bandwidth* go down?
Or should I just have paid more attention when I read the article? :)
Questar - Monday, March 6, 2006 - link
"Needless to say my excitement vanished after I ran the first performance numbers and it offered about half the memory bandwidth of an average Socket-939 platform."jebo - Monday, March 6, 2006 - link
Yeah I haven't been expecting much from the AM-2 move. The way I see it, the move to AM-2 is more a result of DDR prices than performance. It won't be long that similar performing DDR2 will be cheaper than DDR1, so AMD wants to enable it's users to choose the more cost-effective platform. AMD CPU's have never been very memory bandwidth-hungry, so I didn't expect a speed increase. As Anand says, in the future, when we have quad core CPUs and are running several intense background apps/programs at the same time, then bandwidth becomes an issue. As we stand today, we *generally* use somewhere between 1 and 3 threads at a time, and AMD's memory controller simply doesn't need the huge bandwidth that high speed memory offers.DigitalFreak - Monday, March 6, 2006 - link
Now I'm not feeling bad that I just purchased an Opteron 165 to build a 939 system this month. :0)