• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

AMD’s M2 socket, lowering performance?

Cannibal Corpse

Cannibal Corpse

[H]ard|Gawd
Joined
Sep 22, 2002
Messages
1,285
http://www.hardwareanalysis.com/content/topic/56534/

A bite from Anandtech
"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. Remember that the major change with AM2 is the migration to DDR2-667 (and DDR2-800) over DDR-400, so memory bandwidth should go up - not be cut in half."
------------------------------------------------------------

Wow, I am glad that I didn't hold off for this new platform and happily purchased my beloved Opteron 146 + DFI SLI eXpert! Finally I can put my system together! (I finally bought a eVGA 7900GT CO SuperClocked!)
 
Cannibal Corpse said:
http://www.hardwareanalysis.com/content/topic/56534/

A bite from Anandtech
"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. Remember that the major change with AM2 is the migration to DDR2-667 (and DDR2-800) over DDR-400, so memory bandwidth should go up - not be cut in half."
------------------------------------------------------------

Wow, I am glad that I didn't hold off for this new platform and happily purchased my beloved Opteron 146 + DFI SLI eXpert! Finally I can put my system together! (I finally bought a eVGA 7900GT CO SuperClocked!)
Click to expand...
That makes no sense, though... How could bandwidth be half? Then again, since AM2 isn't really out, yet...This is like looking at the Mona Lisa before Divinci finished her face.
 
GilmourD said:
That makes no sense, though... How could bandwidth be half? Then again, since AM2 isn't really out, yet...This is like looking at the Mona Lisa before Divinci finished her face.
Click to expand...

haha, excellent comparison ;)
 
If anything, we will run into what Intel ran into with their switch to DDR2. I know there really wasn't much in the way for improvement there either.
 
Well, it justifies my purchase for a new system! Even 1 year leeway is good enough justification for me!
 
once ddr2-800 becomes the standard it will help out quite a bit. Plus ddr2 is cheaper, and cooler running, which is a huge plus.
 
Chickan said:
once ddr2-800 becomes the standard it will help out quite a bit. Plus ddr2 is cheaper, and cooler running, which is a huge plus.
Click to expand...
I think the point is, though, WTF was Anandtech's staff smoking when they only got half the bandwidth of regular DDR? Incomplete platform or not, sounds like something went wrong.
 
Cannibal Corpse said:
http://www.hardwareanalysis.com/content/topic/56534/

A bite from Anandtech
"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. Remember that the major change with AM2 is the migration to DDR2-667 (and DDR2-800) over DDR-400, so memory bandwidth should go up - not be cut in half."
------------------------------------------------------------

Wow, I am glad that I didn't hold off for this new platform and happily purchased my beloved Opteron 146 + DFI SLI eXpert! Finally I can put my system together! (I finally bought a eVGA 7900GT CO SuperClocked!)
Click to expand...

Seems like the same problem, and time period for the chip, that we saw with Tom's preview of AM2. I think we've agreed that these tests won't represent the final product, and are the result of the signifigant problem with the memory controller. Fixing that problem is likey why AMD moved the launch back from April to June.
 
Gentlemen, it is definitely NOT too early to be drawing conclusions about M2 results. The AM2 Roadmap, published by AMD, states that production on all M2 parts will begin in a little over a month, and parts will be shipped to vendors in about 2 months. If this is not close to what we're going to see in production parts, then AM2 is definitely going to be delayed. It is too close to launch day for AMD to not have a near-finished product.

520_large_amd_m2_timeframe.png
 
Bona Fide said:
Gentlemen, it is definitely NOT too early to be drawing conclusions about M2 results. The AM2 Roadmap, published by AMD, states that production on all M2 parts will begin in a little over a month, and parts will be shipped to vendors in about 2 months. If this is not close to what we're going to see in production parts, then AM2 is definitely going to be delayed. It is too close to launch day for AMD to not have a near-finished product.
Click to expand...

If these chips were brand new, I'd probably agree with you.
But Anandtech said they had their chip in January, Tom's article was from mid/late febuary, meaning their chip probably wasn't fabbed any later than mid January/late either. That gives us 12-14 weeks from when these chips made it to reviewers until production starts, that's time to work things out.
 
I remember toms' and Anandtech said something about a bug in the memery controller. Performance will improve toward launch anyway.
 
Either way, most of us will be moving to Conroe anyway if the Conroe benchies are accurate :).
 
kirbyrj said:
Either way, most of us will be moving to Conroe anyway if the Conroe benchies are accurate :).
Click to expand...
if intel makes a better memory controller i would. until then, on-die efficiency and performance ftw ;)


also, cannibal.. welcome to almost 2 months ago :p
 
To equal...only EQUAL DDR400 @ 7,2,2,2.0 timings you will need DDR2-800 @ 14,4,4,4.0 timings.

If you are talking BH5 you'll need DDR2-1000+ 14,4,4,4.0

Best of luck. Dont confuse AMD's motives for changing with something nessecarily good for YOU the consumer/enthusiast.
 
kirbyrj said:
Either way, most of us will be moving to Conroe anyway if the Conroe benchies are accurate :).
Click to expand...

hehe.. I will never switch back to Intel, never.. I dont care how much better they say or it they really are.. Intel ruined their reputation with me.. sorry Intel :(
 
savantu said:
Hope dies last....
Click to expand...

nope. it never dies ;)

why would AMD release a new socket that performs weaker? it doesnt make sense..

maybe AMD is doing the opposite of Intel... and saying, or having chips available for testing, to perform lower than what it 'really' will.. so when it does come out people are amazed.. and go 'Hey AMD out did themselves.. Good Job!' as opposed to what Intel does and bloats there upcoming benchmarks and then when it comes out everyone is let down.
 
savantu said:
This is with the fixed chips , stepping F , revision F

http://www.computerbase.de/news/har...06/maerz/cebit06_athlon_64_x2_am2_benchmarks/

Hope dies last....
Click to expand...

That's not looking good at all. Did AMD just give up on M2? Seems like they plan to release a mediocre chipset just to appeal to the loyal [read: mindless] AMD followers who will never touch Intel. Meanwhile, Intel is moving forward with rather effective technology, with Conroe, a processor that can beat AMD's best offering in the performance per mhz market, a place AMD once was the proud owner of.

Fortunately, Conroe's gains over the FX-60 aren't as amazing as AMD's Athlon series was against the Pentium 4. So maybe AMD still can get it together, if they don't suddenly shift their focus on K10.
 
Bona Fide said:
That's not looking good at all. Did AMD just give up on M2? Seems like they plan to release a mediocre chipset just to appeal to the loyal [read: mindless] AMD followers who will never touch Intel. Meanwhile, Intel is moving forward with rather effective technology, with Conroe, a processor that can beat AMD's best offering in the performance per mhz market, a place AMD once was the proud owner of.

Fortunately, Conroe's gains over the FX-60 aren't as amazing as AMD's Athlon series was against the Pentium 4. So maybe AMD still can get it together, if they don't suddenly shift their focus on K10.
Click to expand...

AM2 is for quad core. Nothing more nothing less. DDR becomes less forgiving as you start stacking on cores.
 
Bona Fide said:
Gentlemen, it is definitely NOT too early to be drawing conclusions about M2 results. The AM2 Roadmap, published by AMD, states that production on all M2 parts will begin in a little over a month, and parts will be shipped to vendors in about 2 months. If this is not close to what we're going to see in production parts, then AM2 is definitely going to be delayed. It is too close to launch day for AMD to not have a near-finished product.
Click to expand...

You've confused me. Is it definitely not too early, or is it definitely going to be delayed?

Bona Fide said:
http://images.dailytech.com/nimage/520_large_amd_m2_timeframe.png[img][/QUOTE]
Are you supposed to direct-link like that?
Click to expand...
 
HmmmDonut said:
AM2 is for quad core. Nothing more nothing less. DDR becomes less forgiving as you start stacking on cores.
Click to expand...

you, sir, are so devastatingly wrong it's funny.

Socket F is for quad core. Socket AM2 doesn't come close to having the pin count nor integrated capabilities for that.

dave
 
I saw AMD's latest Roadmap today, and I'm not allowed to disclose information (w00t w00t Info Restrictions) but AMD does have something up their sleeve, and they have taken conroe into mind.
 
mikeblas said:
You've confused me. Is it definitely not too early, or is it definitely going to be delayed?
Click to expand...

What I'm saying is a response to the two possibilities.

1. Assuming AMD will launch on time. If we assume that the AMD launch will go as planned (6/6/06), then this M2 sample should be relatively close to the final product. If that is so, then M2 is going to be worse than 939 and definitely worse than Conroe.

2. Assuming the sample is not a final sample. If we assume that the sample given to reviewers was an early test sample, it would mean that the chip is still a long way off from completion. It would also mean that AMD will miss their June 6 launch date.
 
Bona Fide said:
What I'm saying is a response to the two possibilities.

1. Assuming AMD will launch on time. If we assume that the AMD launch will go as planned (6/6/06), then this M2 sample should be relatively close to the final product. If that is so, then M2 is going to be worse than 939 and definitely worse than Conroe.

2. Assuming the sample is not a final sample. If we assume that the sample given to reviewers was an early test sample, it would mean that the chip is still a long way off from completion. It would also mean that AMD will miss their June 6 launch date.
Click to expand...

3. other (unless one works for AMD and more specificly on the AM2 project, I doubt one would know the truth. So 1 and 2 are speculation)
 
Tetrahedron said:
3. other (unless one works for AMD and more specificly on the AM2 project, I doubt one would know the truth. So 1 and 2 are speculation)
Click to expand...
This seems like the most likely answer. Engineering Samples aren't always final designs, and may have significant features disabled or altered from what goes into production.
 
dave_graham said:
you, sir, are so devastatingly wrong it's funny.

Socket F is for quad core. Socket AM2 doesn't come close to having the pin count nor integrated capabilities for that.

dave
Click to expand...

Do you know the difference in the actual pinout between Socket F and Socket AM2?

Socket S1, which has dual channel DDR2, and less pins than Socket 754 (S1 has about 638), will be running a dual core chip (Turion X2). I think pin count isn't the thing to worry about. What the pins are actually doing is what needs to be concerned about.

Main thing I know socket F has so many pins is for add HTT links for more scalability
 
http://www.theinquirer.net/?article=30290
WE HAVE seen many AM2 socket CPUs. Most of them are running at 2.4GHZ and are showcased with DDR 2 800 memory. That is the only way to show any kind of performance difference from the DDR 400 based existing 939 CPUs.
The guys that ran some benchmarks on those machines confirmed that you can expect three to five per cent performance increase and that is about it. AMD claims that those engineering samples are the final revision so you should not expect more of the performance incensement. The existing engineering samples are running at 2.4 GHz but they might end up even faster at the launch date.

However, the price of the DDR 2 will drop and will make it more accessible for the general consumers but DDR 800 wont be anywhere close to mainstream. AMD is playing an interesting game but at least it could support DDR 3 when it comes out, so that is something. µ
Click to expand...

DDR2 gives higher bandwidth--A64 dosen't realy need more.
DDR2 has higher latency--A64 realy dosen't need that.
DDR2 uses less power
DDR3 is more of the same.

AM2=wasted money, get a 939 opti while you still can or better yet wait for Conroe.
Or wait 1H07 for AMD's 65nm. Or Intel's 45nm. Or...
 
mICKEL said:
http://www.theinquirer.net/?article=30290


DDR2 gives higher bandwidth--A64 dosen't realy need more.
DDR2 has higher latency--A64 realy dosen't need that.
DDR2 uses less power
DDR3 is more of the same.
Click to expand...

dual, and eventually quad core chips, need more bandwidth
4-4-4 DDR2 800 (availible now and not hard to find) has the same latency (in time, and therefor in CPU cycles) as 2-2-2 DDR400 which is obviously the fastest you can buy.
yep
and hopefully so

And most importantly:
125W @ 1.2V for the 65nm parts is 104A. Currently S939 boards are probably designed with 90A in mind, early boards, more like 75A. Compatability with many, if not most S939 boards is out the window for high end parts in the next 6 months no matter what socket AMD uses.
 
FreiDOg said:
dual, and eventually quad core chips, need more bandwidth
4-4-4 DDR2 800 (availible now and not hard to find) has the same latency (in time, and therefor in CPU cycles) as 2-2-2 DDR400 which is obviously the fastest you can buy.
Click to expand...
one of these days i need to test latency at 3-3-3, 300mhz vs 2-2-2 at 200mhz and see if that really holds true. i have a strong suspicion that 3-3-3, 300mhz wins the latency battle, as will 4-4-4 at 400mhz :D


Obi_Kwiet said:
/me wants QDR. Or L2 quality SRAM for system memory. That would make me happy.
Click to expand...
hmmm... qdr would be kinda tough to pull off i think.. closest thing to it is XDR, which needs some work imo:
http://www.hardforum.com/showthread.php?t=1031124
 
Alright, something you may not have noticed, but the early AM2 reviews were using a mobo that had no ability to alter memory timings from 4-4-4-12, and they used the board with DDR2-667 ram, not DDR2-800 ram. So yeah, the memory timings used for the reviews were basically a$$. I would expect running tighter timings or higher speed ram would have made a difference.

Listed here under motherboards, the memory timings used.
http://www.tomshardware.com/2006/02/21/a_look_at_amds_socket_am2_platform/page8.html

I haven't seen anyone specifically document either tighter timings or ddr2-800 ram in their feedback about the AM2 platform. I think this is still very much a "wait and see" situation.

Update:
Just an additional bit of info... doesn't look like Intel is going to be having these "wonder" chips available for mass-production until 2007... which is definately long enough for AMD to get some of their AM2 chips out, and revise the process somewhat... then by mid-2007 they'll probably go through a process switch, down to 60nm or whatever is next on the list.

http://www.overclockers.com/tips00932/
 
Obi_Kwiet said:
/me wants L2 quality SRAM for system memory. That would make me happy.
Click to expand...

No you don't. Unless you want to be stuck with 64MB of Main memory again, at the same price that you pay for 1-2GB nowadays.

Then again if Divincis' memery serves him right there are at least 2 spelling and one grammar mistake in this sentence.
 
Logan321 said:
Update:
Just an additional bit of info... doesn't look like Intel is going to be having these "wonder" chips available for mass-production until 2007... which is definately long enough for AMD to get some of their AM2 chips out, and revise the process somewhat... then by mid-2007 they'll probably go through a process switch, down to 60nm or whatever is next on the list.

http://www.overclockers.com/tips00932/
Click to expand...

The article you linked doesn't say that at all. It draws this conclusion from a Daily Tech article:

Nothing unusual about that, those percentages are pretty much standard Intel rampup, but Conroe won't become the mainstream CPU until early 2007

This is talking about market penetration, not availability. When you look at the Daily Tech article, it's pretty clear that Q1 '06 shipments of Pentium D, both Presler and Smithfield, is little more than projected shipments of Conroe in Q3 '06.

The majority of desktop processors sold by Intel are P4 Celerons. Besides, that ball-park estimate for Q3 '06 Conroe shipments is probably more than AMD can ship in Semprons, A64s, and X2 combined. Certainly, one can reference the shipments of Presler + Smithfield from the Daily Tech article against expected shipments of Conroe and come up with a fairly good guess for timeline and availability. But, I'm not going to explain this as it might cause certain people severe mental anguish.

Anyway, since so many people will be upgrading to AM2, that just leaves more Conroe for others. Which is good for those other people. Besides, I'm confident with enough tweaking and quality DDR-2 AM2 will eventually net that 3-5% performance gain.
 
Update:
Just an additional bit of info... doesn't look like Intel is going to be having these "wonder" chips available for mass-production until 2007... which is definately long enough for AMD to get some of their AM2 chips out, and revise the process somewhat... then by mid-2007 they'll probably go through a process switch, down to 60nm or whatever is next on the list.
Click to expand...
are you kidding? Conroe production will begin by early Q3. It'll be available to everyone before end of Q3 :D :D
 
drizzt81 said:
No you don't. Unless you want to be stuck with 64MB of Main memory again, at the same price that you pay for 1-2GB nowadays.
Click to expand...

Probably a lot more than that. There's a reason CPU makers don't put massive amounts of L2 cache in their chips. Actually ther're several reasons, but I digress.
 
FreiDOg said:
dual, and eventually quad core chips, need more bandwidth
4-4-4 DDR2 800 (availible now and not hard to find) has the same latency (in time, and therefor in CPU cycles) as 2-2-2 DDR400 which is obviously the fastest you can buy.
Click to expand...
No, it is in clock cycles, and still lower latency--but more bandwidth. ;) My Gold VX DDR-500 is still better, but a much bigger power consumer. :eek:

I am sure companys like OCZ will make some RAM for AM2 that works better with the ICM than what is avalible today. But Conroe is just plain gona win every witch way there is. AMD will get 65nm going next March(maybe,) and Intel will follow with 45nm. :rolleyes:
 
mICKEL said:
No, it is in clock cycles, and still lower latency--but more bandwidth. ;) My Gold VX DDR-500 is still better, but a much bigger power consumer. :eek:
Click to expand...
i have a feeling 250mhz at 2-2-2 will be pwned by 400mhz at 4-4-4 in bandwidth and latency, check this out:

with the cpu held at 1.8ghz...
200mhz, 2-2-2 = 51.5ns
300mhz, 3-3-3 = 43.8ns

i double checked and made sure i didn't do something odd. interesting eh? ;)
 
You must log in or register to reply here.
Back
Top