Opteron Vs 64 X2

oldbag · 22 Dec 2006 at 19:37

So basically, which one is better, I am assuming the X2 is?

CBP · 22 Dec 2006 at 19:45

Doesn't it depend on which models you're comparing?

eg. opty 144 single core vs 4400 dual core

I went for the opty 170 over the X2 4400 as the opteron models usually gave good overclocking potential.

In my case, with my stepping I was wrong.

jaykay · 22 Dec 2006 at 19:46

yer model numbers need to be included or we have no idea
(mostly an opteron is better if overclocking if not x2 will do)

oldbag · 22 Dec 2006 at 19:48

Oh.. so how about an AMD Athlon 64 X2 Dual Core 4200+ 2.20GHz vs AMD Opteron UP 144 San Diego 1.8GHz 1MB Cache (Socket 939) - Retail

I assume the Athlon would cane it, being dual core?

easyrider · 22 Dec 2006 at 19:57

The 144 is single core the 4200 is dual core.

So you can't compare them like for like.

Tetras · 22 Dec 2006 at 20:18

The 939 Opterons are identical performance to the 939 Athlon 64 or X2 equivalent but some models exist of which there is no equivalent.

Few examples:
Athlon 64 3700+ / Opteron 148 = 2.2 Ghz 1MB cache (single)
Athlon FX-57 / Opteron 154 = 2.8 Ghz 1MB cache (single)

Opteron 165 = 1.8 Ghz 1MB cache per core (dual core)
Athlon X2 4000+ (exists on AM2 socket only) / Opteron 170 = 2.0 Ghz 1MB cache per core (dual core)
Athlon X2 4800+ / Opteron 180 = 2.4 Ghz 1MB cache per core (dual core)
Athlon FX-60 / Opteron 185 = 2.6 Ghz 1MB cache per core (dual core)

All Opterons have 1MB cache per core (so 2MB total for a dual core chip) and are San Diego (Athlon 64) or Toledo (Athlon X2).

The only difference is that they often have nice steppings (that overclock well) and have some models not available with Athlon X2s (such as Opteron 165 which can be had for same price as the X2 3800+, that has only 0.5MB cache per core).

14x, 15x series are single core, 16x, 17x and 18x series are dual core. The Athlon X2 4200+ has no direct equivalent since it is a 512KB per core part but performance would be roughly equal to the Opteron 170 (2.0 Ghz, 1MB per core).

Gashman · 22 Dec 2006 at 22:36

opterons use cherry picked cores don't they? :confused:

Cob · 23 Dec 2006 at 11:51

Gashman said:
opterons use cherry picked cores don't they?

They use cores from nearer the center of the wafer. They require less voltage, thus giving better overclocking results.

Hamoodii · 23 Dec 2006 at 17:00

the opty's overclock better

weescott · 23 Dec 2006 at 17:15

Cob said:
They use cores from nearer the center of the wafer. They require less voltage, thus giving better overclocking results.

So you are saying that cores from the center of a wafer overclock faster? That Opty's only come from the center of a wafer?

Cob · 23 Dec 2006 at 17:55

weescott said:
So you are saying that cores from the center of a wafer overclock faster? That Opty's only come from the center of a wafer?

Cores near the center of the wafer will be of better quality than those nearer the outside. They are used for mobile chips (lower voltage is always needed for laptops) and chips that need to run at the top end of the chart (eg- the AMD FX range).

Opterons don't necessarily come from the center, but a 2.0Ghz Opteron will come from nearer the center than a 2.0Ghz A64. Hence why more 146 Opterons reached around 3.0Ghz than A64 3200+'s. The full 1mb of cache is also more likely to be available and working on cores taken from nearer the centre.

Jokester · 23 Dec 2006 at 18:02

As far as I'm aware there's no variance in "quality" across a wafer, it's the quality of the wafers themselves that determine the performance of the chips and that's determined by all sorts of things, even the day of the week.

Jokester

jaykay · 23 Dec 2006 at 18:09

yer they test the opterons much harder i dont think its which part they come from

A2Z · 23 Dec 2006 at 18:23

i had x2 3800 max went up to was 2.55 with vcore bump, got opty 146 @ 2.8 stable stock volts

aztechnology · 23 Dec 2006 at 21:57

Cob said:
Cores near the center of the wafer will be of better quality than those nearer the outside. They are used for mobile chips (lower voltage is always needed for laptops) and chips that need to run at the top end of the chart (eg- the AMD FX range).

Opterons don't necessarily come from the center, but a 2.0Ghz Opteron will come from nearer the center than a 2.0Ghz A64. Hence why more 146 Opterons reached around 3.0Ghz than A64 3200+'s. The full 1mb of cache is also more likely to be available and working on cores taken from nearer the centre.

If this is true, it's totally new to me. I see no reason why the crystal structure at the edge of the wafers would be worse than the those in the middle.

As far as I'm aware, (and I don't work for AMD), Opterons go through more quality checks and hence are "cherry picked".

How well they clock still depends on the quality of the source wafers, (and that depends on all sorts of things), so you need to look for the good steppings if you want a good clocker. though thats the same for A64's and X2's

Starfall · 23 Dec 2006 at 22:08

Jokester said:
As far as I'm aware there's no variance in "quality" across a wafer, it's the quality of the wafers themselves that determine the performance of the chips and that's determined by all sorts of things, even the day of the week.

Jokester

The centre of the wafer recieves more uniform doping in fabrication, statistically centre dies will be superior to outer dies

aztechnology · 23 Dec 2006 at 22:51

Starfall said:
The centre of the wafer recieves more uniform doping in fabrication, statistically centre dies will be superior to outer dies

My understanding ws that the dopant was introduced in the silicon melt, and whilst there are problems with non-uniformity, I don't think that these are particularly associated with the centre of the wafer. (Though its a long time since I studied all this properly)

Starfall · 23 Dec 2006 at 23:14

aztechnology said:
My understanding ws that the dopant was introduced in the silicon melt, and whilst there are problems with non-uniformity, I don't think that these are particularly associated with the centre of the wafer. (Though its a long time since I studied all this properly)

Chemical vapour deposition and in particular molecular beam epitaxy are not able to impregnate a large substrate evenly accross the whole surface. I didnt think of this before but the lithography processes will also have a larger angle when processing the mask on the outer edges of the wafer.

"in the silicon melt" - The wafer is heated to very high temperatures (600C IIRC) during deposition but it never nears melting temperature...

The processes AMD and Intel use are on the bleeding edge of technology, massively more complex than anything i studied but a lot of the principles will still hold.

locutus12 · 24 Dec 2006 at 01:35

Tetras said:
All Opterons have 1MB cache per core (so 2MB total for a dual core chip) and are San Diego (Athlon 64) or Toledo (Athlon X2).

opterons also come in Denmark core, although i think this may be another name for Toledo,

CBP · 24 Dec 2006 at 11:11

My opteron needs lots of voltage to perform. I probably have the worst stepping available tho