There's a new AM2 preview on Anandtech's page. This looks to be like the almost final version of AM2 and it shows mediocre gains over S939. They're even using DDR2-800. Conroe looks better day by day.
New Am2 Preview on Anandtech's Page
- Thread starter imported_UncivilizedAMD
- Start date
Yeah, well, that's the way it works. I wonder, why didn't they just make a memory controller that was able to use both DDR and DDR2? Let the mobo maker decide what kinda RAM you're gonna use..
It will be interesting to finally see 4-4-4-12 with 2x1Gb off DDR2-800 memory for both platforms by the time Conroe and Windsor are out.
I guess that would increase the die area further still over what they have now, remember Windsor has a die size of 220 mm2, and has better cache density then Toledo did. So more space is taken up by core logic.
It would be bad to support both. Not only would it take up more die space, but it would also hamper the transition to DDR2. It's simply time to forget about DDR and move on. There's really very little point in continued support for it.
Since AM2 doesn't seem to come close to taking advantage of DDRII-800 it seems as if 667MHz or even 533MHz DDRII might be enough to reach the same performance. Those memories should also be available with much tighter timings, which would boost performance by a few percent more. Looks like AM2 is ending up close to where AMD told it would, i.e. a 5% increase in performance. Not that it's very exciting, but I guess one has to be happy for the small things...
It's somewhat strange that AMD hasn't been able to prepare a major architectural overhaul in the three years since the K8 was released. Sure, it is on it's way, but releasing it now would have been very nice, considering the coming of Conroe.
RichUK
Lifer
- Feb 14, 2005
- 10,341
- 678
- 126
No surprise really. Its a dead give away that AMD is not trying to tout a performance increase from its DDR2 mem controller, as the model numbers will stay the same on the equally clocked AM2 and s939 processors.
65nm + their new core (K8L) and 2007 should be AMD's come back against Conroe. Until then s939 should suffice over their current Rev F.
65nm + their new core (K8L) and 2007 should be AMD's come back against Conroe. Until then s939 should suffice over their current Rev F.
I found the preview quite good, though I wish they had tested DDR2 667 just to see if DDR2 800 will really be needed to get that performance advantage over S939. Also, remember that the DDR2 is running at horrendous timings (at least pretty damn bad) while the DDR1 is running at pretty much the best you can get, so I think an overall 5-10% advantage is not too far-fetched.
EDIT: I was under the impression that rev F silicon would actually support both DDR1 and 2 to give AMD enough flexibility over the release schedule (ie, rev F could have been launched for S939). Perhaps this was dropped in the later core revisions or it simply isn't used because it would require more pins.
EDIT: I was under the impression that rev F silicon would actually support both DDR1 and 2 to give AMD enough flexibility over the release schedule (ie, rev F could have been launched for S939). Perhaps this was dropped in the later core revisions or it simply isn't used because it would require more pins.
Well considering for 2x1GB DDR2-800 the best is 4-4-4-12 that you can readily obtain I don't think 4-5-4-15 is completely horrendous. I am talking out of the box settings overclocked timings are another issue entirely.
Unlesss your talking about overclocked timings.... I haven't seen any premium memory brand maker, market DDR2-800 2x1GB modules with 3-3-3-8 timings yet.
A quick searh yeild these results for me:
2x1GB DDR2-533 3-2-2-8 is available from numerous
2x1GB DDR2-667 4-2-2-8 is available from OCZ
I wonder how Windsor will scale though...
n7
Elite Member
- Jan 4, 2004
- 21,281
- 4
- 81
I'm actually impressed that overall, performance is slightly better, or at least not worse.
I'm also glad AT had to test with 1 GB dimms.
Testing with 512 MB dimms is kinda silly now, since the large majority of people get 2x 1024 MB these days.
AMD's A64 architechture is not bandwidth starved, so these results don't surprise me.
I'd assume DDR2-533 @ 3-3-3 or 667 @ tighter timings will be around the same as DDR2-800 @ 4-4-4, etc.
I'm also glad AT had to test with 1 GB dimms.
Testing with 512 MB dimms is kinda silly now, since the large majority of people get 2x 1024 MB these days.
AMD's A64 architechture is not bandwidth starved, so these results don't surprise me.
I'd assume DDR2-533 @ 3-3-3 or 667 @ tighter timings will be around the same as DDR2-800 @ 4-4-4, etc.
Yep. Either AMD has a plan...or they don't. In any case I think everyone knew by now that AM2 wasn't the secret weapon in and of itself.
Has AT said anything about AM2's 4dimms able to do 1T?
anyone know if the revision F mem controller can do that?
If so that's another improvement over s939.
anyone know if the revision F mem controller can do that?
If so that's another improvement over s939.
I'm impressed. Everything will eventually move to DDR2 and AMD is just keeping up with the times. Its taken how long for Intel to come up with something to compete with AMD and AMD is suppose to have an answer to Conroe next year perhaps.
RussianSensation
Elite Member
- Sep 5, 2003
- 19,458
- 765
- 126
It's not surprising. Still using identical A64 architecture besides the redesigned memory controller. If conroe did not obliterate A64 after NetBurst was around since Willamette, Intel shareholders would be flooding the market with sold shares. Frankly, the more Conroe beats A64 the better it is for everyone since A64 squeezed out as much of a performance (esp over P4, P-D) as it could during its reign. Time to move on and let K9/K10 lay the smack down.
As a consumer I'd gladly take 20-30% performance increases over AMD struggling to raise clock speeds every 3 months. Then it'll be Conroe's time to retire. We need loses of performance crown so AMD designs an even better chip next time around and doesn't feed us pure clock speed increases which give 5% boost each performance rating step.
Also despite what most people believe on this forum, selling A64 3200+ chips ($150) in pairs as X2 3800+ ($300) is not necessarily progress as it doesn't let mainstream consumers embrace multi-cores. Certainly $210 conroe chips are a step in the right direction. And until Intel starts to outperform A64 in games, it's hard to imagine gamers adopting dual core processors. Once it does and prices are lower, it's heaven.
I am not even certain if AM2 will stick around or if it's a transitional socket? I thought AMD might adopt DDR3 in Q07 and a new socket to support quad-core cpus too? In contrast 775 conroe seems to be much more forward looking, no?
As a consumer I'd gladly take 20-30% performance increases over AMD struggling to raise clock speeds every 3 months. Then it'll be Conroe's time to retire. We need loses of performance crown so AMD designs an even better chip next time around and doesn't feed us pure clock speed increases which give 5% boost each performance rating step.
Also despite what most people believe on this forum, selling A64 3200+ chips ($150) in pairs as X2 3800+ ($300) is not necessarily progress as it doesn't let mainstream consumers embrace multi-cores. Certainly $210 conroe chips are a step in the right direction. And until Intel starts to outperform A64 in games, it's hard to imagine gamers adopting dual core processors. Once it does and prices are lower, it's heaven.
I am not even certain if AM2 will stick around or if it's a transitional socket? I thought AMD might adopt DDR3 in Q07 and a new socket to support quad-core cpus too? In contrast 775 conroe seems to be much more forward looking, no?
They need to test it with differnt DDR2 speeds. I can't get DDR2 800 for the price of DDR 400 yet. I'm willing to bet that the performance at slower memory speeds could be a few percent less than what we've seen today.
DrMrLordX
Lifer
- Apr 27, 2000
- 22,768
- 12,776
- 136
We'll see more extensive testing when the platform launches on paper or otherwise in May/June or something. Note he couldn't even publish the clock speeds used? It's the best benchmarking we can expect given the circumstances. Anand is passing some useful, albeit slightly veiled information here:
1). Performance gains from first-gen AM2 CPUs will be 5% or lower against a stock s939 CPU of the same speed
2). Many overclocked/enthusiast s939 machines will be as fast as AM2 machines on a clock-per-clock basis
3). Conroe will probably tear AM2 a new one
AM2 might show speed increases over s939 as high as 7-10% with a second core revision, but is anyone really going to sit around waiting for that? Unless they can get a new core revision out by launch, thhhpt!
Bring on AM3 please.
1). Performance gains from first-gen AM2 CPUs will be 5% or lower against a stock s939 CPU of the same speed
2). Many overclocked/enthusiast s939 machines will be as fast as AM2 machines on a clock-per-clock basis
3). Conroe will probably tear AM2 a new one
AM2 might show speed increases over s939 as high as 7-10% with a second core revision, but is anyone really going to sit around waiting for that? Unless they can get a new core revision out by launch, thhhpt!
Bring on AM3 please.
I'm kinda pisses off cuz for many months now, all logic has pointed to AM2 being about 5% faster than s939, and these results bolster that argument. However, for months, i've heard dozens of people insist that AM2's performance will ramp up "significantly" and that we'd all be surprised. I don't think that is happening, and at this rate unless AMD experienced a miracle break through, all those people have been insane.
Seems almost like when Intel went from Northwood to Prescott performance wise, but without the heat/power issues.
Could be a bad year for AMD when Conroe ships.
Could be a bad year for AMD when Conroe ships.
"DDR2-800 at 4-5-4 timings, instead of the 3-3-3 that's possible with this memory on socket 775."
Yes, I know that those timings are pretty much in line with what is currently being officially supported, but by the time AM2 hits the market we should start seeing 3-3-3 modules, while 2-3-2 is the best 1GB DDR1 modules will get. I suppose horrendous is too strong a word but the timings could be better...
Yes, I know that those timings are pretty much in line with what is currently being officially supported, but by the time AM2 hits the market we should start seeing 3-3-3 modules, while 2-3-2 is the best 1GB DDR1 modules will get. I suppose horrendous is too strong a word but the timings could be better...
TRENDING THREADS
-
Discussion Zen 5 Speculation (EPYC Turin and Strix Point/Granite Ridge - Ryzen 9000)- Started by DisEnchantment
- Replies: 25K
-
Discussion Intel Meteor, Arrow, Lunar & Panther Lakes Discussion Threads
- Started by Tigerick
- Replies: 21K
-
Discussion Intel current and future Lakes & Rapids thread- Started by TheF34RChannel
- Replies: 23K
-
-
Facebook
Twitter