Phenome II 920 or 925

[perdomot]

With the AM3 phenoms soon too be released, I was wondering which makes more sense to get. My mobo will handle both AM2+ and AM3 cpus but uses ddr2 ram, not ddr3. Both of these cpus will operate at 2.8Ghz so I'm wondering if I will see any benefit to getting the 925 in performance. Obviously for future upgrades of the mobo/ram it will make sense but will I see any difference on an AM2+ mobo with ddr2 ram?

 

[SunnyD]

The AM3 parts have a slightly faster un-core, meaning it should be a hair faster overall than the 920. Other than they, they are supposedly the same other than the DDR3 support.

 

[DrMrLordX]

Couldn't you bump up the nb speed (I didn't think Phenoms had an "uncore", isn't that a Core i7 thing?) to bring a 920 up to 925 speeds? If the 925s are binned better and OC better then I can see an advantage to them, but still . . .

 

[SunnyD]

Originally posted by: DrMrLordX
Couldn't you bump up the nb speed (I didn't think Phenoms had an "uncore", isn't that a Core i7 thing?) to bring a 920 up to 925 speeds? If the 925s are binned better and OC better then I can see an advantage to them, but still . . .

You could, but that affects everything else too. As far as uncore, it's Intel's term really, but it's been expanded into AMD territory. AMD's reference clock is their HT Link.

I would expect the 925 to be a bit better overall due to later yields rather than first-rans.

 

[Zstream]

Originally posted by: SunnyD

Originally posted by: DrMrLordX
Couldn't you bump up the nb speed (I didn't think Phenoms had an "uncore", isn't that a Core i7 thing?) to bring a 920 up to 925 speeds? If the 925s are binned better and OC better then I can see an advantage to them, but still . . .

You could, but that affects everything else too. As far as uncore, it's Intel's term really, but it's been expanded into AMD territory. AMD's reference clock is their HT Link.

I would expect the 925 to be a bit better overall due to later yields rather than first-rans.

No, HT has always been HT....

 

[SunnyD]

Originally posted by: Zstream

Originally posted by: SunnyD

Originally posted by: DrMrLordX
Couldn't you bump up the nb speed (I didn't think Phenoms had an "uncore", isn't that a Core i7 thing?) to bring a 920 up to 925 speeds? If the 925s are binned better and OC better then I can see an advantage to them, but still . . .

You could, but that affects everything else too. As far as uncore, it's Intel's term really, but it's been expanded into AMD territory. AMD's reference clock is their HT Link.

I would expect the 925 to be a bit better overall due to later yields rather than first-rans.

No, HT has always been HT....

Right... but in this "modern era" where most everybody forgot that AMD even makes processors, it's now commonly (incorrectly) referred to as uncore.

 

[masteryoda34]

The AM3 CPU's are rumored to overclock somewhat better.

 

[Cookie Monster]

Un-core speeds of 2.0GHz (compared to 1.8GHz)
TDP down to 65W from 95W
DDR3 and DDR2 support (i.e can be used on AM2+ or AM3 boards)
Overclock maybe ~200MHz better according to leaked testing over at XS

Not sure if it supports HT3.1 though.

 

[Idontcare]

Originally posted by: Cookie Monster
Un-core speeds of 2.0GHz (compared to 1.8GHz)
TDP down to 65W from 95W
DDR3 and DDR2 support (i.e can be used on AM2+ or AM3 boards)
Overclock maybe ~200MHz better according to leaked testing over at XS

Not sure if it supports HT3.1 though.

Do we know for certain that the higher uncore and lower TDP are not specific to AM3 chips operating in AM3 mode with DDR3?

In other words, how certain are we that an AM3 PhII isn't setup internally to "revert back" to the lower uncore clockspeed and operate at a higher TDP (different IMC circuitry and voltage is used for DDR2 interface vs. DDR3) when popping the chip into an AM2/AM2+ mobo?

Up till know all I've seen is a lot of assumption about this topic but none have been qualified by any technical reasons as to why or how AMD is suddenly going to be able to increase the uncore clockspeed and drop the TDP on AM3 chips when the chips are put into AM2+ sockets. Doesn't add up for me.

 

[Idontcare]

Originally posted by: Cookie Monster
Un-core speeds of 2.0GHz (compared to 1.8GHz)
TDP down to 65W from 95W
DDR3 and DDR2 support (i.e can be used on AM2+ or AM3 boards)
Overclock maybe ~200MHz better according to leaked testing over at XS

Not sure if it supports HT3.1 though.

Do we know for certain that the higher uncore and lower TDP are not specific to AM3 chips operating in AM3 mode with DDR3?

In other words, how certain are we that an AM3 PhII isn't setup internally to "revert back" to the lower uncore clockspeed and operate at a higher TDP (different IMC circuitry and voltage is used for DDR2 interface vs. DDR3) when popping the chip into an AM2/AM2+ mobo?

Up till know all I've seen is a lot of assumption about this topic but none have been qualified by any technical reasons as to why or how AMD is suddenly going to be able to increase the uncore clockspeed and drop the TDP on AM3 chips when the chips are put into AM2+ sockets. Doesn't add up for me.

 

[Flipped Gazelle]

Originally posted by: Idontcare

Originally posted by: Cookie Monster
Un-core speeds of 2.0GHz (compared to 1.8GHz)
TDP down to 65W from 95W
DDR3 and DDR2 support (i.e can be used on AM2+ or AM3 boards)
Overclock maybe ~200MHz better according to leaked testing over at XS

Not sure if it supports HT3.1 though.

Do we know for certain that the higher uncore and lower TDP are not specific to AM3 chips operating in AM3 mode with DDR3?

In other words, how certain are we that an AM3 PhII isn't setup internally to "revert back" to the lower uncore clockspeed and operate at a higher TDP (different IMC circuitry and voltage is used for DDR2 interface vs. DDR3) when popping the chip into an AM2/AM2+ mobo?

Up till know all I've seen is a lot of assumption about this topic but none have been qualified by any technical reasons as to why or how AMD is suddenly going to be able to increase the uncore clockspeed and drop the TDP on AM3 chips when the chips are put into AM2+ sockets. Doesn't add up for me.

About a month ago my contact at AMD told me that AM3 CPU's will attain the lower TDP only with AM3/DDR3.

 

[Cookie Monster]

Double post from IDC.. BLASPHEMY!

Anyway, the newer 9x5 models could use lower voltage (The PII X4 910 tested at XS uses 1.175v which is pretty low, and god knows why AMD named this AM3 processor "910"), hence a different revision compared to the initial denebs. But I guess your right if the chip itself is relatively the same, and that the lower TDP only applies to AM3/DDR3 and not AM2+/2.

Everything that i listed was based on the roadmap released by AMD, and in the ATs "deneb" review. So you can take it with a pinch of salt since these aren't 100% official.

 

[Idontcare]

Originally posted by: Cookie Monster
Double post from IDC.. BLASPHEMY!

Anyway, the newer 9x5 models could use lower voltage (The PII X4 910 tested at XS uses 1.175v which is pretty low, and god knows why AMD named this AM3 processor "910"), hence a different revision compared to the initial denebs. But I guess your right if the chip itself is relatively the same, and that the lower TDP only applies to AM3/DDR3 and not AM2+/2.

Everything that i listed was based on the roadmap released by AMD, and in the ATs "deneb" review. So you can take it with a pinch of salt since these aren't 100% official.

Ha ha, yeah getting my reply's to post with the AT server today has been a crapshoot...not sure if I am the only one experiencing it but about 50% of my attempts to reply end with a "oops that link no longer works" response from the server.

I actually assumed my post to this thread was dead, I didn't bother to check. Now I see not only was it not dead but the forum server felt it worthy of double-posting to maximize your viewing pleasure.

At any rate thanks for the response and letting me know where the info came from, if it comes from AT article then it is good enough for me.

 

[VirtualLarry]

Yeah, the server has been rather spotty today.

 

[DrMrLordX]

Originally posted by: SunnyD

Right... but in this "modern era" where most everybody forgot that AMD even makes processors, it's now commonly (incorrectly) referred to as uncore.

Actually, the AMD equivalent to Intel's uncore is their northbridge, or nb. All K10-class chips with on-die L3 cache have an on-die northbridge that governs the speed of the L3 cache and the integrated memory controller. Whether or not the L3-deprived Rana, Propus, and Regor chips have an on-die northbridge, I do not know, but that's another matter altogether.

The nb speed itself may be internally represented in the form of a multiplier factored against the HT speed, but for the purpose of chip specifications, nb speed is nb speed. No mention of HT is/was made in reference to nb speed.

But I'm just splitting hairs here, really.

Originally posted by: Idontcare

Do we know for certain that the higher uncore and lower TDP are not specific to AM3 chips operating in AM3 mode with DDR3?

In other words, how certain are we that an AM3 PhII isn't setup internally to "revert back" to the lower uncore clockspeed and operate at a higher TDP (different IMC circuitry and voltage is used for DDR2 interface vs. DDR3) when popping the chip into an AM2/AM2+ mobo?

Up till know all I've seen is a lot of assumption about this topic but none have been qualified by any technical reasons as to why or how AMD is suddenly going to be able to increase the uncore clockspeed and drop the TDP on AM3 chips when the chips are put into AM2+ sockets. Doesn't add up for me.

I too share your skepticism. I can see the possibility of the integrated memory controller running at a different voltage in an AM3 socket, possibly due to a different reference voltage. As we all know, the amount of power pumped through the memory controller on AMD chips is affected by vdimm, so the default 1.8v vdimm of DDR2 will not only consume more power on the DIMM side of things but should, at least theoretically, cause the memory controller to consume more power as well.

Whether or not I buy a .3v difference to the IMC accounting for a 30W difference in TDP is an entirely different story. It may be that the entire chip boots with a different vid depending on the socket.

Interestingly enough, there will be AM3 boards tht support DDR2 memory. One wonders if AM3 chips will still be rated as 65W TDP processors when used in boards such as this. Also . . . I had thought that a majority of the AM2+ Deneb-based X4s were 125W TDP processors and not 95W? Or am I off-base here? Will the 125W processors become 95W processors in their AM3 incarnations?