Sign in with
Sign up | Sign in
Your question

AT AM2 preview, not at all hopeful

Last response: in CPUs
Share
April 10, 2006 11:04:43 AM

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=273...
It looks bad. The memory controler is bandwith limited. We aren't seeing the bandwith increase that we should but we are suffering from higher LCs. Thus the performance increase is dissapointing.
All AMD lovers, wait untill 2007 when 65 nm is implemented if you are looking for a DDR2 switch.
That leaves Conroe being the only solution left for a real performance increase. Of course that chip may fail too though until now it's looking good.
Eh, Inq. weren't jocking. 3-5 % that tends to be a lot less in some apps.
Now what's the real problem apart from minor increases? Well put it simply, price. DDR2 800 will cost a LOT and so will the first AMD CPUS. People who like AMD should better stick to their 939.

More about : am2 preview hopeful

a b à CPUs
April 10, 2006 11:10:00 AM

At least now, there is indeed a small boost in gaming framerates with DDR2, much better than the deficit we saw two months ago...

However....

I agree. COnroe will be the DDR2 platform
April 10, 2006 12:19:52 PM

Given that AMD was supposed to launch in June at Computex, the fact that AM2 was performing so poorly just five months before launch was cause for worry.
Even with DDR2-800 at the best timings we could manage back then, Socket-AM2 was unable to outperform Socket-939 at DDR-400.


Hopefully they will fix what ails it, but with so much to do it seems that any performance increases will have to wait for 65nm. Why release it at all if it isn't faster than a 939 on DDR400 ? They are going to supply the Intel Fanboys with so much cannonfodder it ain't funny.
If this is so, I'm going Conroe, I don't want to have to buy 2 CPU's within 4 or 5 months, that's just plain stupid.
AMD........what's wrong with you guys all of a sudden ?
Related resources
Can't find your answer ? Ask !
April 10, 2006 12:35:30 PM

We can point this out. Intel had the same problem other cpus. let take a look at the frist P4 vers p3. Also Northwood .130nm to .090nm Presscott. So Both Intel and Amd has there ups and downs. Which helps US.
April 10, 2006 12:43:26 PM

I just hope they both can keep a even speed give or take 5% Which helps both Amd and Intel.
April 10, 2006 12:56:40 PM

Quote:

All AMD lovers, wait untill 2007 when 65 nm is implemented if you are looking for a DDR2 switch...DDR2 800 will cost a LOT and so will the first AMD CPUS. People who like AMD should better stick to their 939.


How will 65nm remove a memory controller bandwidth bottleneck? Oh, I want it to - I'm an AMD fanboy! I just was wondering how 65nm will acheive this.

2007 suits me fine. My grand plan has been to leapfrog S939 in about next March - and jump straight from good ole socket A [XP3200+] to AM2 [cheap initial processor w/ overclockability] ... dang, though, a whole entire year of drooling in anticipation (and time to switch fan allegiance if I have to)
April 10, 2006 12:59:38 PM

That would be great, provided it's, not rushed to market with problems like past Intel and AMD CPUs, which is good for no one.
I saw a post that a guy was testing his Conroe with no fan running on the heatsink and it finished Prime, and only got to 27C. Well if that's the case, I won't be waiting till Q1 2007 for AMD.
Let the Beacon of Performance Light The Way ( and I don't care who makes it)
April 10, 2006 1:24:54 PM

It will increase the number of transistors. And AMD will probably get working on tweaking that memory controller cause right now things aren't looking too shabby.
April 10, 2006 1:38:30 PM

Quote:
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=273...
It looks bad. The memory controler is bandwith limited. We aren't seeing the bandwith increase that we should but we are suffering from higher LCs. Thus the performance increase is dissapointing.
All AMD lovers, wait untill 2007 when 65 nm is implemented if you are looking for a DDR2 switch.
That leaves Conroe being the only solution left for a real performance increase. Of course that chip may fail too though until now it's looking good.
Eh, Inq. weren't jocking. 3-5 % that tends to be a lot less in some apps.
Now what's the real problem apart from minor increases? Well put it simply, price. DDR2 800 will cost a LOT and so will the first AMD CPUS. People who like AMD should better stick to their 939.





Anand let me down on this one. Everyone knows that 4-5-4 will not be great for AMD IMC. I would have thought he would do the same thing that he did for the recent 955 review. The RAM he used will do like 3-2-3 at 667. I'd like to see a latency vs. speed analysis.
April 10, 2006 1:44:32 PM

Quote:
Given that AMD was supposed to launch in June at Computex, the fact that AM2 was performing so poorly just five months before launch was cause for worry.
Even with DDR2-800 at the best timings we could manage back then, Socket-AM2 was unable to outperform Socket-939 at DDR-400.


Hopefully they will fix what ails it, but with so much to do it seems that any performance increases will have to wait for 65nm. Why release it at all if it isn't faster than a 939 on DDR400 ? They are going to supply the Intel Fanboys with so much cannonfodder it ain't funny.
If this is so, I'm going Conroe, I don't want to have to buy 2 CPU's within 4 or 5 months, that's just plain stupid.
AMD........what's wrong with you guys all of a sudden ?



Everyone knows that Intel go tno increase from the first DDR2. As better latency comes out (there's still two months left) there will be more improvements. Besides AMD hasthe best position right now. They don't have to lower prices for Conroe. It will take until 2007 for Conroe to really saturate the market because there are millions of PDs and P4s cooling their heels in Dell's warehouses. It would hurt Intel more financially to just leave the overstock hanging so Conroe may be a bad thing for the bottom line. I still think that AM2 will get 20% before the year is out (without 65nm). I would bet AMD is at work on the next minor revision. They have to release intermediate revisions before Rev G.
April 10, 2006 1:50:11 PM

Quote:
Hopefully they will fix what ails it, but with so much to do it seems that any performance increases will have to wait for 65nm. Why release it at all if it isn't faster than a 939 on DDR400 ? They are going to supply the Intel Fanboys with so much cannonfodder it ain't funny.
If this is so, I'm going Conroe, I don't want to have to buy 2 CPU's within 4 or 5 months, that's just plain stupid.
AMD........what's wrong with you guys all of a sudden ?

Agree. There's not much AMD can change within the two month period given the time stocking is needed.
What we see today is a 99% certainty of what we'll see on June 6th.

One can only pray now that AMD brings 65nm out ahead of schedule to close/overtake the performance gap Conroe has created.





Conroe doesn't exist yet. AM2 is not out yet. Conroe will hurt Intel cause they'll be stuck with cheap dual cores Dell won't even be able to give away. I would say it's closer to 70% of the final because the speed/latency of RAM will improve by June. I'm sure mobo makers are trying to get the fastest timings out of their BIOS'. AMD would have to have tested with low latency lower speed RAM. Let's just wait and see.
April 10, 2006 1:58:13 PM

There is no way AM2 will get those kinds of improvements. 20%? If they pass 5% is enough considering this. Bandwidth won't do much, while latencies maybe but considering that lower lcs decrease band. it won't be much either.
April 10, 2006 2:19:52 PM

Alright, I've actually read the article now.. and I have to say, I didn't find it to be nearly as negative as some of you. Though I have admitted being an AMD fanboy, I don't think it's me that's being obsessive. More importantly, I might not be as technically minded as some of you.

From the article, we can see what socket AM2 isn't - it's not the next generation of AMD processor innovation. All it is, is an update of the pretty good current s939 Athlon64 range to use DDR2 RAM, because DDR2 RAM is now cheaper, and because maybe after doing an unbelievably good job with DDR1 for so long, AMD finally needs to arrange headroom for its future processors.

You could say they could have waited for those future processors before changing socket, but now's a good enough time to begin not forcing new buyers to buy obsolete RAM. I had a socket A board where I could use either SDR RAM or DDR RAM - but since Athlon64 has its integrated memory controller, I guess they can't arrange that flexibility so easily. So, from now on, all new current generation Athlon64's will be dedicated to the new RAM.

Remember, the early AM2's were slower than equivalent s939, but now they're slightly faster - so it's equal or better.

So for you guys who want to dismiss AM2, I say ... fair enough!

* If you have an s939 system - I'd just stick to that as long as you can, so you don't need to update your RAM until later when it's cheaper and better. You can either keep your system or upgrade your processor, and it might even be cheaper to upgrade only the processor now, and do the whole mainboard/processor/RAM thing later even if it's only 6 months down the track.

* If you don't have an s939 system and you love the sound of Conroe (just wait and see how the real thing is, rather than relying on onstage demos and preview reviews) - well, if Conroe's as good as it's cracked up to be, go for it! :p  You might want to consider the long term upgrade path - I'm no expert, but I think Intel has a history of changing their sockets a lot, though it's been good since LGA775 hasn't it? (With AMD changing more after sktA, I suppose! esp. the broken promise of a long s939) But maybe it's alright because it's so good. [Conroe]

hnmmm ... yep, that's it. That's how I see it as of now.
April 10, 2006 2:29:58 PM

Quote:
Conroe doesn't exist yet.

Yes it does. I've seen it benchmarked with my own two eyes, and it is hitting near world record marks @ 2.4GHz, 65W TDP, $309. See Wusy's post above for an example.

Quote:
I'm sure mobo makers are trying to get the fastest timings out of their BIOS'. AMD would have to have tested with low latency lower speed RAM. Let's just wait and see.

The difference here is DDR2 has existed for years and years. When Intel made the transition, it was relatively new technology. They knew they needed it to stay even somewhat competitive with AMD's dominance through their integrated emory controller that was much tighter than the Northbridge solution at the time. You think that MoBo munfacturers haven't been trying their hardest to get the best poerformance out of DDR2 for years? They were just saving up their ideas to implement them in the next two months? Grow up, this is pure fanboy fantasy.

The facts are: AM2 is AMD's addition of support for DDR2 memory, and that is about it. It is not "next-generation" anything like many folks were hoping. It doesn't really affect performance at all (for now) while adding cost to the platform. Hopefully it will see performance inprovement in the future, but it would be unrealistic to expect more than a 5-7% change over the next couple months.

Conroe very much exists and even with pre-release leaked engineering samples is mopping the floor with everything that you can currently buy, from either Intel or AMD. Intel knows this and they know that is may lead to inventory issues, which is why they are having a huge price decrease on their existing chips in May. Hopefully this will force AMD to lower their prices as well, and consumers win either way.

As time passes some companies have more competitve products than others. It is cycle. Intel looks to have a very very dominant position at least until 2007, and we'll see what happens then.

-iterations
April 10, 2006 2:43:25 PM

It looks to me that AMD are trying a different route than Intel. AM2 looks like to be an update to the current AMD processors, especially the whole move to DDR2. Of course, the performance increase isn't special, but these are 90nm processors, and 65nm could be different and bring the performance boost that so many want to see AMD provide. Personally, I'm just gonna hold out and see how it all turns out, because if Intel does regain the price/performance ratio crown, then back to Intel I go... if not, I'll just be sticking to AMD then... pure and simple.
April 10, 2006 2:55:56 PM

You'll be surprised when AMD busts out with DDR3.
April 10, 2006 3:01:51 PM

Quote:
You'll be surprised when AMD busts out with DDR3.


Yes it's gotta be better, it has a "3" in it, and DDR2 only has a "2" in it....

I hope this was a joke.
April 10, 2006 3:02:02 PM

Does the SuperPI test mean anything? Yes, my X2 4200+ is slow with only 41s on 1M, but with only 57% CPU usage on average. What gives?
April 10, 2006 3:08:44 PM

Means SuperPI is not a threaded app... and so is not taking advantage of both cores to produce a decent score.

And MrsD, concerning DDR3, I would have to see the result of that, though the even bigger bandwidth would be good, I wonder what the latencies would be like though....
April 10, 2006 3:21:48 PM

SuperPI does utilize both cores as shown by windows task manager. My second SuperPI 1M run even shown only with 50% CPU usage.

The 2.4Ghz Conroe runs at 3.072v??? That's an amazing high voltage for a CPU with only 65W TDP even at 65nm. How do intel manage to achieve that?
April 10, 2006 3:28:17 PM

SuperPi is generally used by Overclockers and enthusiasts to rank their systems by pure number crunching ability. It is popular because using the same workload (calculating Pi) but changing one parameter (the number of decimal places) you can exercise deeper and deerper portions of the memory hierarchy.
April 10, 2006 3:29:33 PM

Quote:
SuperPI does utilize both cores as shown by windows task manager. My second SuperPI 1M run even shown only with 50% CPU usage.

The 2.4Ghz Conroe runs at 3.072v??? That's an amazing high voltage for a CPU with only 65W TDP even at 65nm. How do intel manage to achieve that?


That version of CPUId doesn't properly support Conroe, the real voltage is 1.18-1.35v

-iterations
April 10, 2006 3:59:01 PM

Quote:
It looks bad. The memory controler is bandwith limited. We aren't seeing the bandwith increase that we should but we are suffering from higher LCs. Thus the performance increase is dissapointing.

What a dork you are. This quote is from the chart Anand posted:
Quote:
ScienceMark 2.0 (Bandwidth) 5007 MB/s 6805 MB/s 36%
ScienceMark 2.0 (Latency 512-byte stride) 53.83 ns 49.77 ns 7.5%


Quote:
We'll start at the bottom of the table and go up from there. Rev F processors feature a 128-bit DDR2-800 memory controller, which works out to offer a peak theoretical bandwidth to/from memory of 12.8GB/s. As you can expect, that's twice the bandwidth of Rev E CPUs' 128-bit DDR-400 controller at 6.4GB/s. Thus to see a 36% increase in memory bandwidth according to ScienceMark is to be expected, albeit a bit on the low side.


Quote:
What's even more important than the increase in memory bandwidth is that access latency has been reduced by 7.5% over the DDR-400 memory controller in the Rev E cores. Lower latency and more bandwidth means that, at bare minimum, performance won't go down

http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2738&p=4

Rev G and H will reach the full bandwidth of DDR2-800. Even Anand states that AMD still has 2 months to tweek even more the core.
Quote:
The verdict is also not out on Rev F; although the launch is only two months away, we keep on hearing that availability won't be until July. While that's not enough time for AMD to be making major changes to the silicon, it is quite possible that the changes have already been made and they're just waiting to get new chips back from the fab
.
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2738&p=6

Ohh, and this is what he has to say about Conroe:
Quote:
However times are changing, and after a very long hiatus Intel will soon resume increases in FSB frequency, not to mention that their new Core architecture is considerably more data hungry than anything we've seen to date.
8)

So once more, who is bandwidth limited?? :wink:
April 10, 2006 6:36:44 PM

You're right about one thing. Yup i was wrong, lower LCs but still distant timings.
And please know that a chip that takes advantage of bandwidth will with no doubt take advantage of DDR2 memory.
April 10, 2006 7:31:02 PM

Quote:

It looks bad. The memory controler is bandwith limited. We aren't seeing the bandwith increase that we should but we are suffering from higher LCs. Thus the performance increase is dissapointing.


Sorry but the analysis on this is just plain wrong. Per the article the AMD CPU's do not suffer from a lack of memory bandwith as evidence by previous bandwith increases. With a 35% gain in bandwith there is little increase in performance which means that at this time bandwith is not a primary issue (if the CPU had been begging for more bandwith you would have seen a jump in performance). The access latency per the article is actually less than the DDR by 7.5% in the test measuring it, so the higher timings probably are not a huge factor (especially since they should b elower than what was able to be used in testing per the article)

The biggest thing to realize is that AM2 processors are the same processor. Just because a new socket has been introduced does not mean the processor is going to be any better especially having seen Intel switch to DDR2 with the same lack of performance gain. DDr2 is meant to solve future problems rather than address a current need. So a 3800 X2 with AM2 socket is basically a 3800 x2 939 part with the inherent problems of DDR2 that Intel has also encountered.

That being said AMD will need to introduce new processors to compete with Conroe, either the frequency will have to be increased or significant revision done to keep ahead of Intel.
April 10, 2006 7:36:34 PM

Probably i missinterpret bandwidth limited. By that i mean it doesn't need bandwidth,thus more of it won't bring much performance. Sry, i always mess up these terms.
April 10, 2006 8:17:22 PM

Quote:

One can only pray now that AMD brings 65nm out ahead of schedule to close/overtake the performance gap Conroe has created.


I guess that they are not really in a rush to keep that old architecture alive. Yes, they brought DDR2 support for marketting and availability reason, but AM2 wont and will never be the best thing AMD will bring against Conroe.

Keeping an old architecture alive too long hurts.. look at Intel's netburst. It should have ben scrapped with Prescott. But they kept it for a reason. To have time to get something NEW out. Just like AMD is probably doing.. Not wasting too much time on current k8, simply adapt it to newer market, and developping something better for after Conroe.
April 10, 2006 8:57:47 PM

Quote:
That BS is just as bad as some of MMMs.

Conroe doesn't exist yet.

Head over to XtremeSystems and see the Conroe(ES) yourself with provided benchmarks.

Quote:
I would say it's closer to 70% of the final because the speed/latency of RAM will improve by June

Stocking takes months not days.


Ramping isn't the key, the key is how many P4s are and PDs are left. IF Dell has to decide, there will be millions of P4s left in the warehouse, if they get rid of the P4s then they can't sell Conroe in major volume.

GET IT? Conroe is BAD for Intels bottom line.
April 10, 2006 9:00:11 PM

Quote:
There is no way AM2 will get those kinds of improvements. 20%? If they pass 5% is enough considering this. Bandwidth won't do much, while latencies maybe but considering that lower lcs decrease band. it won't be much either.



Then I'll be wrong. Should I wear a chicken suit or something?
April 10, 2006 9:16:54 PM

Quote:
* If you have an s939 system - I'd just stick to that as long as you can, so you don't need to update your RAM until later when it's cheaper and better. You can either keep your system or upgrade your processor, and it might even be cheaper to upgrade only the processor now, and do the whole mainboard/processor/RAM thing later even if it's only 6 months down the track.

* If you don't have an s939 system and you love the sound of Conroe (just wait and see how the real thing is, rather than relying on onstage demos and preview reviews) - well, if Conroe's as good as it's cracked up to be, go for it! You might want to consider the long term upgrade path - I'm no expert, but I think Intel has a history of changing their sockets a lot, though it's been good since LGA775 hasn't it? (With AMD changing more after sktA, I suppose! esp. the broken promise of a long s939) But maybe it's alright because it's so good. [Conroe]




I'm planning on upping my 4400+ to an FX60 so I won't be buying Conroe OR AM2 when they come out. I just know how AMD behaves in a heavily used system like mine and am confident that FX62 won't be a joke.
April 10, 2006 9:33:18 PM

i think it is a case of wait and see the final finished CPU from each of Intel and AMD then and only then can a true meaningful comparison be made until then keep consulting the crystal balls it makes for a good read.

At the end of the day it doesn't matter who is in front of the other in the market aslong as at the end of the day they continue to push each other forward bettering the technology as they go we the consumer will benefit so its win win :lol: 
April 10, 2006 11:34:33 PM

I'm ashamed.....my thoughts of defection are reprehensible, and quite premature.
!