AM2 X2 5000+

[JAG87]

if the FX62 has 400 HT, slap some DDR2-800 in there and watch out Intel. Wow, 333 was something to hope for, but if AMD makes it 400...

 

[Suspicious-Teach8788]

Originally posted by: DeathBUA

Originally posted by: n7
Hmm wow, i just rechecked this older roadmap
http://www.c627627.com/AMD/Athlon64/

I think it actually might just be right too.

Anyone notice how the FX-62 is a 7x400 CPU according to that? :Q

If that's actually accurate, wow.
Huge jump in HTT speed if that's actually gonna happen!

If that older roadmap is correct then OCing might not be so horrid, take that 4200+ dualie, 333x7, if you can get the HTT up to 400 then you're at 2.8 and FX-62 levels. Not too shabby eh? Thats only a 67 point hop in HTT which could be doable even on the earlier revisions and Mobo's

That roadmap was good in the days of 2004, 2005, but today, it is out of date. No one even updates that site anymore...

BTW no way was that a Feb 2005 update. It was a Feb 2006 update. I'm almost certain because I saw it updated almost every few days back in 2005 up until fall where they had less to fix

 

[Mogadon]

Originally posted by: DLeRium

Originally posted by: DeathBUA

Originally posted by: n7
Hmm wow, i just rechecked this older roadmap
http://www.c627627.com/AMD/Athlon64/

I think it actually might just be right too.

Anyone notice how the FX-62 is a 7x400 CPU according to that? :Q

If that's actually accurate, wow.
Huge jump in HTT speed if that's actually gonna happen!

If that older roadmap is correct then OCing might not be so horrid, take that 4200+ dualie, 333x7, if you can get the HTT up to 400 then you're at 2.8 and FX-62 levels. Not too shabby eh? Thats only a 67 point hop in HTT which could be doable even on the earlier revisions and Mobo's

That roadmap was good in the days of 2004, 2005, but today, it is out of date. No one even updates that site anymore...

BTW no way was that a Feb 2005 update. It was a Feb 2006 update. I'm almost certain because I saw it updated almost every few days back in 2005 up until fall where they had less to fix[/]

That roadmap may be old, but the newer roadmaps indicate the AM2 chips as having an HTT of 200 whilst someone who actually has a chip has confirmed the HTT is at 333 ... which has to make you wonder.

 

[n7]

Originally posted by: Mogadon
[BTW no way was that a Feb 2005 update. It was a Feb 2006 update. I'm almost certain because I saw it updated almost every few days back in 2005 up until fall where they had less to fix[/]

That roadmap may be old, but the newer roadmaps indicate the AM2 chips as having an HTT of 200 whilst someone who actually has a chip has confirmed the HTT is at 333 ... which has to make you wonder.

Yeah, that's updated from 2006 i'm sure.

I didn't even notice it said 2005 there.

I'm starting to think that "old" roadmap might just be right, or at least right for the higher end models like the 5000+
Maybe lower end models will be 200, other 333?

All i know is it'll be interesting to see.

One thing i found interesting is that AT specifically said nothing about native HTT speeds in their AM2 preview
Could be cause they aren't allowed to maybe

 

[JAG87]

EXACTLY, neither did Tom.

you figure they get an ES chip, that they would give you more detailed info, like HT and multiplier, because I am sure that people are interested in knowing how to overclock these new chips, but NOBODY said anything on HT speeds and multipliers. I am starting to get a feeling that AMD requested for that information not be released and sent a crippled cpu, just to screw up the preview benchmarks and make people think that AM2 is nothing better then 939. just watch them pull out a chip that outperforms 939 by a large gap...

 

[KeithTalent]

Originally posted by: n7

Originally posted by: Mogadon
[BTW no way was that a Feb 2005 update. It was a Feb 2006 update. I'm almost certain because I saw it updated almost every few days back in 2005 up until fall where they had less to fix[/]

That roadmap may be old, but the newer roadmaps indicate the AM2 chips as having an HTT of 200 whilst someone who actually has a chip has confirmed the HTT is at 333 ... which has to make you wonder.

Yeah, that's updated from 2006 i'm sure.

I didn't even notice it said 2005 there.

I'm starting to think that "old" roadmap might just be right, or at least right for the higher end models like the 5000+
Maybe lower end models will be 200, other 333?
All i know is it'll be interesting to see.

One thing i found interesting is that AT specifically said nothing about native HTT speeds in their AM2 preview
Could be cause they aren't allowed to maybe

I bet you are right:

Roadmap

AMD Athlon 64 X2 (Windsor) is expected to be released on June 6th for the Socket AM2 platform on a 90nm SOI process. Windsor is a dual core CPU featuring AMD's Presidio Security technology and Pacifica Virtualisation technology. Windsor will introduce a new Socket AM2 platform (940 pins) and is expected to feature DDR2 800 support and contain either a 2Mb (2 x 1MB) or 1MB (2 x 512KB) L2 cache. Windsor is expected to feature better power management to previous steppings and lower power consumption through the optimisation of transistor leakage currents. The fastest initial version is the 5000+ which features a clock speed of 2.667Ghz running on a 333Mhz FSB (8x333Mhz).

AMD Athlon 64 (Orleans) is expected to be released on June 6th for the Socket AM2 platform on a 90nm SOI process. Orleans is the cut down version of Windsor, featuring a single core design with 1Mb L2 cache alongside the features of Windsor such as DDR2 667, Presidio Security technology and Pacifica Virtualisation technology. Like Windsor, it will be built for the Socket M2 platform.

Should be a fun summer

 

[Furen]

Originally posted by: n7
Yeah, that's updated from 2006 i'm sure.

I didn't even notice it said 2005 there.

I'm starting to think that "old" roadmap might just be right, or at least right for the higher end models like the 5000+
Maybe lower end models will be 200, other 333?

All i know is it'll be interesting to see.

One thing i found interesting is that AT specifically said nothing about native HTT speeds in their AM2 preview
Could be cause they aren't allowed to maybe

Ok, Just so you guys know having a higher HTT base clock will not yield better performance at all. If you wanna test it just benchmark your CPUs at 200x9 and then at 300x6. Performance should pretty much be the same if you also use a memory divider to keep the memory from being overclocked.

AT didnt specifically say "HTT will be 200MHz" but in the Overclocking section of the AM2 preview they ran the BOTH THE AM2 and the S939 chips at 200x10 and 250x10. I suppose they could have an unlocked CPU but I think they would have run it at the stock HTT if only to make the benchmarking more accurate.

By the way, I highly doubt we'll see Z-ram this year. Next year may be a different story but consider that the licensing happened a few months ago and validating chips takes quite a long time...

 

[IEC]

and to think that I was just considering upgrading to socket 939...

 

[nycdude]

wow. Interesting stuff from AMD. :Q

 

[cmrmrc]

but if 5000+ is 8x333=2667,
then that means that 4800/4600 are 7x333=2333,
4400/4200 are 6x333=2000, and
4000/3800 are 5x333=1667....
unless they make some with .5 multiplier, but if so, theres will only be a 166mhz speed bump for every two models...sounds weird to me...

 

[Furen]

I was under the impression that rev F would allow half multipliers, and I'm sure we'll at least see this on Semprons.

 

[Oemo]

Some more food for thought, the 5000 has a date of 0603. Means it was made in the third week of this year. If the OP is right then AMD has kept the 333HT under wraps since atleast January.

 

[jiffylube1024]

Why is everybody talking about the HTT speed like it matters so much? 266HTT is what I think AM2 is going to be for starters, but it really doesn't matter if it's 266 or 333. It's not like the P4, which needs FSB bandwidth; AMD uses an adjustable HTT multiplier anyways.

 

[jiffylube1024]

Originally posted by: Oemo
Some more food for thought, the 5000 has a date of 0603. Means it was made in the third week of this year. If the OP is right then AMD has kept the 333HT under wraps since atleast January.

Nothing surprising about that. They've probably had alpha silicon for a year now. Plus there's nothing conclusive about the 333HT, and it's nothing to jump up and down about even if true...

If legit (which I'm sure it is), it's from a test production run, which companies have to do to make sure they can manufacture the product before they actually launch it.

Remember that Conroe setup that Anand tested a month ago? That was obviously manufactured in February at the latest, yet Conroe doesn't actually launch for 5 more months.

 

[the Chase]

Originally posted by: jiffylube1024

Originally posted by: Oemo
Some more food for thought, the 5000 has a date of 0603. Means it was made in the third week of this year. If the OP is right then AMD has kept the 333HT under wraps since atleast January.

Nothing surprising about that. They've probably had alpha silicon for a year now. Plus there's nothing conclusive about the 333HT, and it's nothing to jump up and down about even if true...

If legit (which I'm sure it is), it's from a test production run, which companies have to do to make sure they can manufacture the product before they actually launch it.

Remember that Conroe setup that Anand tested a month ago? That was obviously manufactured in February at the latest, yet Conroe doesn't actually launch for 5 more months.

This is true. It might not make any difference in performance at all. And this may or may not be true. I don't know the OP and we have only seen the IHS of the chip and no other verification.

But being he actually has the chip I'll take his word on the specs, which is what makes this so interesting.....

For quite a while now the 5000+ was supposed to be 2.6Ghz with 2 x 512KB of L2 cache. Now close to launch we find out it may be 2.66Ghz with 2 x 1MB of cache. Enough of a difference to make me wonder if there is more that we don't know about what is coming with AM2 chips.....Who knows - fun to speculate.

 

[Fox5]

Originally posted by: JAG87
EXACTLY, neither did Tom.

you figure they get an ES chip, that they would give you more detailed info, like HT and multiplier, because I am sure that people are interested in knowing how to overclock these new chips, but NOBODY said anything on HT speeds and multipliers. I am starting to get a feeling that AMD requested for that information not be released and sent a crippled cpu, just to screw up the preview benchmarks and make people think that AM2 is nothing better then 939. just watch them pull out a chip that outperforms 939 by a large gap...

Paranoid much?

Ok, Just so you guys know having a higher HTT base clock will not yield better performance at all. If you wanna test it just benchmark your CPUs at 200x9 and then at 300x6. Performance should pretty much be the same if you also use a memory divider to keep the memory from being overclocked.

Bah, don't use a divider, since the dawn of overclocking, it's always been important to keep the memory in sync with the bus to achieve max performance!

 

[jiffylube1024]

Originally posted by: Fox5
Bah, don't use a divider, since the dawn of overclocking, it's always been important to keep the memory in sync with the bus to achieve max performance!

Ever since AMD's onboard memory controller, which drastically cuts latencies, you no longer need to keep the memory in synch with the HT clock.

Heck, the A64 doesn't even use a true traditional FSB!

 

[Fox5]

Originally posted by: jiffylube1024

Originally posted by: Fox5
Bah, don't use a divider, since the dawn of overclocking, it's always been important to keep the memory in sync with the bus to achieve max performance!

Ever since AMD's onboard memory controller, which drastically cuts latencies, you no longer need to keep the memory in synch with the HT clock.

Heck, the A64 doesn't even use a true traditional FSB!

Just because the impact is lessened doesn't mean it isn't there.

 

[Mogadon]

Originally posted by: Fox5

Originally posted by: jiffylube1024

Originally posted by: Fox5
Bah, don't use a divider, since the dawn of overclocking, it's always been important to keep the memory in sync with the bus to achieve max performance!

Ever since AMD's onboard memory controller, which drastically cuts latencies, you no longer need to keep the memory in synch with the HT clock.

Heck, the A64 doesn't even use a true traditional FSB!

Just because the impact is lessened doesn't mean it isn't there.

Just because it's there it doesn't mean it matters in real world applications.

 

[Fox5]

Originally posted by: Mogadon

Originally posted by: Fox5

Originally posted by: jiffylube1024

Originally posted by: Fox5
Bah, don't use a divider, since the dawn of overclocking, it's always been important to keep the memory in sync with the bus to achieve max performance!

Ever since AMD's onboard memory controller, which drastically cuts latencies, you no longer need to keep the memory in synch with the HT clock.

Heck, the A64 doesn't even use a true traditional FSB!

Just because the impact is lessened doesn't mean it isn't there.

Just because it's there it doesn't mean it matters in real world applications.

In that case you have no reason to be concerned about AM2, or any athlon 64 beyond a 2800+, since the performance won't really matter in most real world applications.
Come on, AM2 is only giving a 2-5% increase over 939, that's the same kind of performance increase that can be gotten out of slightly playing wiht memory settings, and most people say it doesn't matter.

 

[Mogadon]

Originally posted by: Fox5

Originally posted by: Mogadon

Originally posted by: Fox5

Originally posted by: jiffylube1024

Originally posted by: Fox5
Bah, don't use a divider, since the dawn of overclocking, it's always been important to keep the memory in sync with the bus to achieve max performance!

Ever since AMD's onboard memory controller, which drastically cuts latencies, you no longer need to keep the memory in synch with the HT clock.

Heck, the A64 doesn't even use a true traditional FSB!

Just because the impact is lessened doesn't mean it isn't there.

Just because it's there it doesn't mean it matters in real world applications.

In that case you have no reason to be concerned about AM2, or any athlon 64 beyond a 2800+, since the performance won't really matter in most real world applications.
Come on, AM2 is only giving a 2-5% increase over 939, that's the same kind of performance increase that can be gotten out of slightly playing wiht memory settings, and most people say it doesn't matter.

Whatever man to compare the memory latency issue to being the same as the difference between a low end CPU and a high end CPU is ridiculous. The area where i'd say I push my hardware the most is gaming and when you're running a CPU intensive game there is a world of difference between an A64 2800+ and an A64 running at 2.8GHz, trust me! Much more than the difference gained by running memory in sync, the days of overclocking AthlonXP's where dividers are highly detremental to performance are past dude, catch up ... .

I never said the 5% didn't matter, and I was being facetious when I said you wouldn't notice it in the real world, you will, it'll be the difference between having 47 fps and 52 fps or those extra 10 seconds you'll save when rar'ing a large file. At the moment that extra performane gain isn't worth the extra $100 or so bucks, to me.

Also, part of the point about this thread is the discussion regarding whether AM2 is only going to give us a 3-5% increase or if AMD have something extra to pull out of the hat. The whole memory latency thing has been covered in many, many, many threads past, go find yourself one of those to play in rather than taking this off topic.

 

[Ika]

um... so if AM2's HTT speed is higher than 200, would that mean increased performance...?

 

[Acanthus]

Originally posted by: guoziming
um... so if AM2's HTT speed is higher than 200, would that mean increased performance...?

no

 

[Duvie]

Someone likened the AMD HTT link as being a 6 lane highway while older FSB style were 3 lane highways....The fact of the matter is if they are not clogging 6 lanes up then adding more lanes to it will add nothing in terms of performance...

When the chips went from 800x2 or 1600 (earlier sckt 754/939 chips) to 1000x2 or 2000 of the 90nm chip winchester we saw no tangible increases....

 

[Furen]

Originally posted by: Fox5
Bah, don't use a divider, since the dawn of overclocking, it's always been important to keep the memory in sync with the bus to achieve max performance!

It is impossible to run a K8 with a 1:1 memory controller to DRAM ratio since the memory controller runs at CPU speed. The memory, thus, always runs on a divider. Hypertransport plays no part in this since the hypertransport is not between the CPU and the memory controller like regular front-side buses are.