AMD Athalon 64 X2 wattage choices

[IMGordy2U]

I noticed that some X2 processors come in 2 different wattages -- 65W vs 89W. I would assume that the lower wattage proc would run cooler, but what do you give up for that? Why would anyone want the higher wattage proc over the lower wattage proc, assuming all the other specs are the same?

 

[djplanet]

Higher wattage allows better performance. Thus the higher clocked processors will be in the 89 range. All other specs being the same, yes the lower wattage would be better (with heat dissipation, easier on the power bill, etc.), but might cost more (I could be wrong, not entirely sure).

 

[IMGordy2U]

I'm referring to AMD Athalon 64 X2 procs that are the same model (like 4200+ & 4600+), but have a 65W and a 89W version, both having the same clock speed out of the box.

 

[darkstar782]

The lower wattage ones are cooler and better overclockers but cost more

 

[Action_Man]

The lower wattage ones are cooler and better overclockers but cost more

Actually, they are not better overclockers, reviewers have given it pretty much thumbs down on OCing.

Jack

This is true.

 

[darkstar782]

Thats wierd, you'd have thought they would have been binned as better performing parts and by increasing them to the 'standard' voltage/wattage they would perform better.

How come that isnt true? Is it some intentional limitation on AMD's part?

 

[IMGordy2U]

Check here for more info from AMD.

Apparently there is no difference, except the cost to make the lower wattage variation more energy efficient. Now, how they can do that without sacraficing anything else is still a mystery.

 

[darkstar782]

I'm still confused as to why that is. All the Engergy efficient examples are (we are told) is the same CPU but rated for a lower Vcore.

Either:

1. They just dropped the Vcore on some equally-binned dies, in which case the standard ones should also run fine at the lower vcore this would explain the lack of extra overclocking headroom, but means AMD are being decietful imho.

2. They used higher-binned parts, dropped the Vcore, and reduced the speed to that of a lower binned part. This explains why there are no EE versions of the top cores - there were no higher binned parts availible, but in this case they should overclock better :s

I'm probably missing something vital here, hopefully Jack will post a detailed explaination, then it wont bug me anymore

 

[evilr00t]

I think they just dropped the default Vcore on the CPUs. When I read AnandTech's Athlon64 EE article, they specified the voltages and speed at which each EE A64 ran at.

Processor power consumption can be modeled as being porportional to Frequency * Voltage * Voltage, and I found that the EEs weren't very far off the thoeretical model. So, AMD doesn't seem to be using a different process to get the EE CPUs.

1. They just dropped the Vcore on some equally-binned dies, in which case the standard ones should also run fine at the lower vcore this would explain the lack of extra overclocking headroom, but means AMD are being decietful imho.

I suspect they are pulling higher binned parts and running them at a lower default voltage.

Take my Prescott256 as an example: my Prescott256 can clock up to 3.7 easily on 1.4V (stock cooling!!). On the other hand, I can run the Prescott256 at 1.0V at 2.4GHz, where it only outputs 1/3 of the heat.

 

[darkstar782]

I think they just dropped the default Vcore on the CPUs. When I read AnandTech's Athlon64 EE article, they specified the voltages and speed at which each EE A64 ran at.

Processor power consumption can be modeled as being porportional to Frequency * Voltage * Voltage, and I found that the EEs weren't very far off the thoeretical model. So, AMD doesn't seem to be using a different process to get the EE CPUs.

1. They just dropped the Vcore on some equally-binned dies, in which case the standard ones should also run fine at the lower vcore this would explain the lack of extra overclocking headroom, but means AMD are being decietful imho.

I suspect they are pulling higher binned parts and running them at a lower default voltage.

Take my Prescott256 as an example: my Prescott256 can clock up to 3.7 easily on 1.4V (stock cooling!!). On the other hand, I can run the Prescott256 at 1.0V at 2.4GHz, where it only outputs 1/3 of the heat.

Exactly, but if they are using higher-binned parts they should be better overclockers, and people are saying they are not, so I'm curious as to why

 

[Slobogob]

I think they just dropped the default Vcore on the CPUs. When I read AnandTech's Athlon64 EE article, they specified the voltages and speed at which each EE A64 ran at.

Processor power consumption can be modeled as being porportional to Frequency * Voltage * Voltage, and I found that the EEs weren't very far off the thoeretical model. So, AMD doesn't seem to be using a different process to get the EE CPUs.

1. They just dropped the Vcore on some equally-binned dies, in which case the standard ones should also run fine at the lower vcore this would explain the lack of extra overclocking headroom, but means AMD are being decietful imho.

I suspect they are pulling higher binned parts and running them at a lower default voltage.

Take my Prescott256 as an example: my Prescott256 can clock up to 3.7 easily on 1.4V (stock cooling!!). On the other hand, I can run the Prescott256 at 1.0V at 2.4GHz, where it only outputs 1/3 of the heat.

Exactly, but if they are using higher-binned parts they should be better overclockers, and people are saying they are not, so I'm curious as to why

They use a different manufacturing process - at least that´s what i think. Otherwise those chips would overclock better, waaaay better. The difference can´t be too complex though, that would drive the prices through the roof.

 

[evilr00t]

Exactly, but if they are using higher-binned parts they should be better overclockers, and people are saying they are not, so I'm curious as to why Smile

From what I recall, Anand got their EE chip to 2.8 on air, which is not an "amazing" result compared to non-EE chips.

The reason they're not good (by "good", I mean "better than normal A64X2s") overclockers is probably because they don't have to pull from much higher bins (probably only one step speed grade up) to get the processor working at a lower voltage. Most of AMD's current A64X2s can run at 2.4 at non-EE voltages anyway (for a good example of this, the good 'ol 3800+ to 5000+ OC comes to mind), so taking the bottom bin and running it at 1.8@1.025V / 2.2@1.2V shouldn't be too difficult. You'll notice that AMD's EE chips don't run faster than 2.2 GHz.

Stolen from Anandtech forums, post by JumpingJack:

Anand,
Viewing the specs for the EE vs non-EE they appear to by undervolting the EEs to achieve the lower power, and likely down-binning but it is not clear. Here is the comparision I would like to see:

EE vs non-EE at identical clocks and Vcore across a matrix, say take a 3800+ do a matrix:

2.0 GHz @ 1.20, 1.25, 1.30, 1.35
2.2 GHz @ 1.20, 1.25, 1.30, 1.35
etc.
Put it under full load and measure power and temperature (use the same thermal solution at the same fanspeed RMP), disable cool and quiet.

My suspicion is that they are doing more than simply undervolting and that there is a processing trick that is getting an extra 10%. For example, at 1.20 the dynamic power would only scale down about 15%, or roughtly 89 TDP to 75 TDP, another 10 watts needs to come from somewhere -- is it binned from frquency, is it a leakage trap/trick, could these be just super low yielding straing 90 SOI wafers?

Doing the skew above on both a non-EE and EE would allow one to determine where within the matrix of bin's AMD may be choosing the EE processors and why we have such low availability.

Thanks.

JumpingJack: The reason the TDP might have decreased may be because AMD's 90nm process has improved since its debut; EE chips only exist on AMD's more recent 90nm process.

...As for where the AnandTech article is, I don't know where it is, it's disappeared!

 

[evilr00t]

For example, at 1.20 the dynamic power would only scale down about 15%, or roughtly 89 TDP to 75 TDP, another 10 watts needs to come from somewhere

So, how do they make up the last 10W?

 

[darkstar782]

No I don't think this is the case, if this were case then their entire product line would have moved down the power scale and they would triumphantly declare power victory. These are either special binned CPUs (less likely now that I see the data) or simply underclocked/undervolted CPUs put aside and rebranded from higher bins (like say the 2.6 GHz pile that should make 5000+ ).

Right now the data is point to simply undervolting and underclocking, that is a huge power savings. But the binning needs to start high to be affective, which explains the relatively scarcity of the EEs.

Jack

So, in that case, why dont the EEs overclock like the higher-binned chips would be expected to?

I don't doubt that you are 100% correct, I just hate things that dont make sense to me

 

[m25]

True, but I'd not be surprised of the inverse; just more stable chips. 90nm K8s undervolt pretty well (my A64 is an example with stable 1.20V operation and there are a lot like that). It could be that they noticed this capability after consolidating the process and just made a differentiation.
Dropping to 1.20V from 1.40 lowers my load temps ~5-6°C and the difference will be even wider for dual cores; they are less efficient in dissipating heat.