I am currently building a PC with an eye to using it as a media centre, but keeping the power draw as low as possible.
I had been planning on putting an AMD 4450e processor in it, because I have seen a few people who have managed to get a complete system down to around 20W by undervolting that processor.
However, I have just noticed that the gruntier 4850e (2.5Ghz as opposed to 2.3Ghz) is only a few pounds more.
My question is, presuming I underclock the 4850e to the same speed I was planning on underclocking the 4450e to, would I be able to get the same idle power draw out of it as they are both built on the same architecture with (I presume) the same TDP.
Thanks!
AMD 4450e / 4850e - Same Undervolting?
Moderators: NeilBlanchard, Ralf Hutter, sthayashi, Lawrence Lee
Someone once told me that all three of the CPU's in the line (including the 4050e) are actually the same chip just clocked differently. And that AMD decides which speed to clock it by testing the chips (the ones that test better at the faster speeds are clocked faster). Supposedly this is the same for almost all CPU's (not just these particular AMD chips), and most memory chips also (Value memory is set to run a bit slower or have slower recommended timings because it does not test as well at higher speeds).
I don't know whether the above explanation is actually true, but it makes sense because it seems like it would be costly to actually make so many different designs with such a small difference in clock speed.
BTW, I have the 4050e in a Linux server that I use for software development and I use the stock clock speeds (2.1 Ghz). Cool n Quiet is enabled. This is one amazing CPU in that I am able to run it at idle at about 2 C over ambient room temperature with a passive Xigmatek 1283 HS, although do I have it in a Antec Mini P180 with the 200mm top fan not too far from the HS running at slow speed (3-speed TriCool Fan).
I was so concerned about getting the thermal paste (AS Ceramique) to spread out and cure properly, that I ran CPU burn for several hours when I first built the system.
I don't know whether the above explanation is actually true, but it makes sense because it seems like it would be costly to actually make so many different designs with such a small difference in clock speed.
BTW, I have the 4050e in a Linux server that I use for software development and I use the stock clock speeds (2.1 Ghz). Cool n Quiet is enabled. This is one amazing CPU in that I am able to run it at idle at about 2 C over ambient room temperature with a passive Xigmatek 1283 HS, although do I have it in a Antec Mini P180 with the 200mm top fan not too far from the HS running at slow speed (3-speed TriCool Fan).
I was so concerned about getting the thermal paste (AS Ceramique) to spread out and cure properly, that I ran CPU burn for several hours when I first built the system.
-
austinbike
- Posts: 192
- Joined: Sun Apr 08, 2007 5:09 pm
Basically speaking, when companies make chips, a typical wafer is going to have all the same design on each die.
Because of impurities and minute fluctuations in the design process, each of the dies is unique. Not like snowflakes, but close.
When the dies are tested (still on the wafer), you'll see that they can yield a multitude of speeds. They cut the wafer apart and each of the dies is embedded in a package.
They go through the test, sort and marking, each becoming a particular speed of chip. They may be 2.4GHz, 2.3GHz or even 2.2GHz even though they are next to each other on the wafer.
Even though the die may yield a 2.4GHz processor, if the company has a lot of demand for 2.2's and less for 2.4's, they may end up marking and fusing the chip as 2.2, not 2.4. That is why some processors can be easily overclocked. However, that does NOT mean that buying a 2.2 would allow you to save money and overclock to a 2.4, because that die may only have yielded a 2.2 and won't overclock stably.
When people talk about how certain processors are better than others in overclocking, what they are really saying in most cases is that THEIR chip overclocks well.
When you see chips offered in standard power, low power and high performance, it is all the same design and layout, they just yield differently. Sometimes the cache size is different, but that is usually just disabling some of the cache lines vs. a discrete design.
Clock speed, volting, TDP and other characteristics are all settings that are fused on the processor in the production process, they are not characteristics that are designed in.
Because of impurities and minute fluctuations in the design process, each of the dies is unique. Not like snowflakes, but close.
When the dies are tested (still on the wafer), you'll see that they can yield a multitude of speeds. They cut the wafer apart and each of the dies is embedded in a package.
They go through the test, sort and marking, each becoming a particular speed of chip. They may be 2.4GHz, 2.3GHz or even 2.2GHz even though they are next to each other on the wafer.
Even though the die may yield a 2.4GHz processor, if the company has a lot of demand for 2.2's and less for 2.4's, they may end up marking and fusing the chip as 2.2, not 2.4. That is why some processors can be easily overclocked. However, that does NOT mean that buying a 2.2 would allow you to save money and overclock to a 2.4, because that die may only have yielded a 2.2 and won't overclock stably.
When people talk about how certain processors are better than others in overclocking, what they are really saying in most cases is that THEIR chip overclocks well.
When you see chips offered in standard power, low power and high performance, it is all the same design and layout, they just yield differently. Sometimes the cache size is different, but that is usually just disabling some of the cache lines vs. a discrete design.
Clock speed, volting, TDP and other characteristics are all settings that are fused on the processor in the production process, they are not characteristics that are designed in.
The highest clocked/lowest voltage chips are generally the top bin and the best for overclocking or undervolting. Pretty much it is luck of the draw for how good any particular chip is. I bought two "G2" stepping X2s recently a 4400+ and a 5000+. The 4400+ happened to be able to undervolt a few steps below the 5000+ at the same clockspeed.chrisps wrote:I guess that does make sens, and I guess that if it can just handle higher clock speeds, then there is no reason why it shouldnt under clock just as well.
Can anyone confirm this?
When you can buy mobile and desktop cpus for the same socket the mobile chips tend to be the best overclockers/undervolters.
The 1.15V 4850e should have the best chance to be the best chip. I don't think my 4400+ can run 2.5GHz on 1.15V, but it is close.
