So this is Larrabee?

I'm sorry, I edited my last post. I love the numbers from Company of Heroes, less power consumption with twice the fps count. Can we consider the performance difference now? ;)

In the previous test, it was X3100 vs X1250.

This test is rather skewed because they use a special low-power 4850e CPU, which they pit against a regular E7200. The 4850e has a 20W TDP advantage by default. So it says little about the IGP itself. In fact, if we add 20W to the AMD numbers, Intel would come out the clear winner.
 
We are getting off topic now but I can't let this go just yet without telling the truth.

AMD 4850e:
2.5GHz
65nm
1.15V - 1.25V
45W

Intel E7200:
2.53GHz
45nm
0.85V - 1.3625V
65W

In the test, both CPUs are running with 1.25V.
 
We are getting off topic now but I can't let this go just yet without telling the truth.

AMD 4850e:
2.5GHz
65nm
1.15V - 1.25V
45W

Intel E7200:
2.53GHz
45nm
0.85V - 1.3625V
65W

In the test, both CPUs are running with 1.25V.

So? The 4850e still has a 20W TDP advantage, as your numbers clearly show.
Even if we were to assume that there's a linear relationship between voltage and power dissipation, the Intel would still have a TDP of (65/1.36)*1.25 = ~60W, still giving the AMD a 15W advantage.
That difference alone is probably more than the TDP of the IGP itself, therefore drawing any conclusions about the IGP seems incredibly far-fetched.
 
So? The 4850e still has a 20W TDP advantage, as your numbers clearly show.
Even if we were to assume that there's a linear relationship between voltage and power dissipation, the Intel would still have a TDP of (65/1.36)*1.25 = ~60W, still giving the AMD a 15W advantage.
That difference alone is probably more than the TDP of the IGP itself, therefore drawing any conclusions about the IGP seems incredibly far-fetched.

LOL!!!!

Looks like the "professsional software developer" forgot about a simple formula: P = V^2 / R :eek:
 
LOL!!!!

Looks like the "professsional software developer" forgot about a simple formula: P = V^2 / R :eek:

I didn't forget, I just said to assume a linear equation for simplicity's sake.
But if you insist:
P = V^2 / R
65 = 1.85 / R
65R = 1.85
R = 0.028

So the new P would be P = 1.56 / 0.028 = 55.8W.

Doesn't really change the story much, AMD's CPU still has a bigger TDP advantage than what is reflected in the measured results. So still the chipset/IGP in the Intel system is probably more power-efficient than the AMD one.
 
I didn't forget, I just said to assume a linear equation for simplicity's sake.
But if you insist:
P = V^2 / R
65 = 1.85 / R
65R = 1.85
R = 0.028

So the new P would be P = 1.56 / 0.028 = 55.8W.

Doesn't really change the story much, AMD's CPU still has a bigger TDP advantage than what is reflected in the measured results. So still the chipset/IGP in the Intel system is probably more power-efficient than the AMD one.

The number I posted before was after a BIOS update. Before the update, the numbers look much worse:
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=3288
 
Well, if a BIOS update affects power consumption, that is obviously a factor that is not part of the hardware design.

Why don't you ask your friend pxc about how AMD is giving out a lower TDP value than the actual value used by their CPU and how Intel is giving out a higher TDP value than what their CPUs actually use. I can find it for you if the search function is still working.

Btw are you telling me that a 2.5GHz 65nm AMD CPU is using less power than a 2.5GHz 45nm Intel CPU? I thought that you said AMD is struggling with their 65nm production.
 
Why don't you ask your friend pxc about how AMD is giving out a lower TDP value than the actual value used by their CPU and how Intel is giving out a higher TDP value than what their CPUs actually use. I can find it for you if the search function is still working.

Btw are you telling me that a 2.5GHz 65nm AMD CPU is using less power than a 2.5GHz 45nm Intel CPU? I thought that you said AMD is struggling with their 65nm production.

Why is there such strong cognitive dissonance in your head when Intel IGPs are mentioned?
 
Back to the topic, eventhough Intel is a good CPU manufacturer with technological advantages over their rivals, this doesn't mean that they can easily be a good GPU manufacturer too.

They are still far behind both AMD and nVidia in term of performance, features and power consumption. Their IGP is the prove which shows that Intel need to do a lot more work to be in the same playing field with nVidia and AMD.

Unless Intel have showed us some real numbers which make the Larrabee competitive against nVidia's and AMD's offerings, I don't think that it is appropriate to suggest that Larrabee will be a good product based on Intel performance in the CPU business alone.
 
Unless Intel have showed us some real numbers which make the Larrabee competitive against nVidia's and AMD's offerings, I don't think that it is appropriate to suggest that Larrabee will be a good product based on Intel performance in the CPU business alone.

You've pretty much been suggesting the opposite in this entire thread, which is equally inappropriate.
 
Looks like the "professsional software developer" forgot about a simple formula: P = V^2 / R :eek:
You would probably want to start off with P = f x C x V^2 (switching transistor power), but AMD's 65nm transistors and Intel's 45nm transistors have very different characteristics. You can't compare them so easily, certainly not by voltage alone and more importantly, each have different numbers of transistors.

The easiest way to compare power usage would be to look at how much power each one consumes running the same heavy load(s). Some sites have tested CPU power consumption from the 12V aux connector. The loads used may not represent the highest you can get, but it's at least comparable.

alg7_munif said:
Why don't you ask your friend pxc about how AMD is giving out a lower TDP value than the actual value used by their CPU and how Intel is giving out a higher TDP value than what their CPUs actually use.
Which AMD power figure would you like: the underrated APC, "max" (usually quoted by AMD as "Wattage") or the embarassingly high "TDP"? I didn't say all AMD processors have underrated power usage, but AMD is clearly playing with the numbers to make several processors seem much more power efficient than they are. If you doubt how overrated Intel 45nm chips power ratings are, just ask your good friend Charlie @ theinq: http://www.theinquirer.net/gb/inquirer/news/2007/03/29/intel-talks-specifically-about-penryn
Word on the street is that Intel is fudging the TDP numbers up, they could be pulling the 65W parts down to 45-50W and still not be fibbing. The rest of the TDPs would go down accordingly. Basically, these chips are a lot lighter on real world power use than Intel is claiming.
Which is actually what is seen in real testing. I wonder if you have compared a Penryn heatsink to an AMD stock 65W heatsink. :p Hint: Intel's 45nm 65W cooler is a lot smaller than AMD's 65W cooler.
 
You would probably want to start off with P = f x C x V^2 (switching transistor power), but AMD's 65nm transistors and Intel's 45nm transistors have very different characteristics. You can't compare them so easily, certainly not by voltage alone and more importantly, each have different numbers of transistors.

The easiest way to compare power usage would be to look at how much power each one consumes running the same heavy load(s). Some sites have tested CPU power consumption from the 12V aux connector. The loads used may not represent the highest you can get, but it's at least comparable.

Which AMD power figure would you like: the underrated APC, "max" (usually quoted by AMD as "Wattage") or the embarassingly high "TDP"? I didn't say all AMD processors have underrated power usage, but AMD is clearly playing with the numbers to make several processors seem much more power efficient than they are. If you doubt how overrated Intel 45nm chips power ratings are, just ask your good friend Charlie @ theinq: http://www.theinquirer.net/gb/inquirer/news/2007/03/29/intel-talks-specifically-about-penryn
Which is actually what is seen in real testing. I wonder if you have compared a Penryn heatsink to an AMD stock 65W heatsink. :p Hint: Intel's 45nm 65W cooler is a lot smaller than AMD's 65W cooler.

The frequency is similiar for both processors so you can leave that as constant and capacitance is an inverse function of impedance. What I'm trying to say is the difference of power consumption between the 4850e and E7200 is not as big as what Scali2 suggested.

I'm sure that the E7200 has a lower, if not much lower consumption than the suggested 65W TDP on the spec because it is a 45nm part. The 4850e might have 45W "max wattage" but I think that in real world both of them will have more or less the same amount of power consumption.

Please don't tell me you believe that the 65nm 4850e uses 20W less power than the undervolted 45nm E7200 in Anand's article.
This test is rather skewed because they use a special low-power 4850e CPU, which they pit against a regular E7200. The 4850e has a 20W TDP advantage by default. So it says little about the IGP itself. In fact, if we add 20W to the AMD numbers, Intel would come out the clear winner.
 
I'm sure that the E7200 has a lower, if not much lower consumption than the suggested 65W TDP on the spec because it is a 45nm part. The 4850e might have 45W "max wattage" but I think that in real world both of them will have more or less the same amount of power consumption.
Yeah, probably.
 
No, Intel puts more cache on their chips because they have a larger transistor budget than AMD. The amount of transistors that Intel spends on cache is larger than the amount that AMD spends on cache + memory controller.
Nehalem still has more cache than any AMD processor.

I know this is a late reply, but dont make shit up out of your own ass. It has been stated that is the exactly the reason why they have larger caches, because accessing the cache is still much, much faster than accessing ram without an on die memory controller.
 
I know this is a late reply, but dont make shit up out of your own ass. It has been stated that is the exactly the reason why they have larger caches, because accessing the cache is still much, much faster than accessing ram without an on die memory controller.

You seem to have broken logic.
Cache is much faster than accessing ram, period.
An on-die memory controller might be marginally faster than a chipset, but it's still MUCH slower than cache.
Anyone who thinks you don't need cache because you have an on-die memory controller is an idiot.
AMD seems to agree with me, because they're going from 2 MB L3 cache to 6 MB. Much closer to Intel-like cache sizes than before. Their problem was just that such a large cache wouldn't be economically viable with their 65 nm manufacturing process. With 45 nm it is.
 
Larger cache always helps. But the real reason intel's core 2 quads need more cache is because they're essentially two duo's glued together on a single die. The only way for cores 1-2 to work with 3-4 is to use the larger cache. Nehalem will change this with die similarities to AMD's phenom and further improves on that design. Going a bit off topic now.
 
Larger cache always helps. But the real reason intel's core 2 quads need more cache is because they're essentially two duo's glued together on a single die. The only way for cores 1-2 to work with 3-4 is to use the larger cache.

Not really.
Cores 1 and 2 can communicate through the shared L2 cache.
Cores 1/2 and 3/4 can only communicate through the FSB. In that case, the size of the cache doesn't really matter. Even without any cache they could still communicate via FSB.

In fact, Nehalem and Phenom use L3 cache *specifically* to improve core-to-core communication.
So what you say doesn't make sense.
 
I agree with some of the other posters in this thread about the drivers. I have confidence that Intel can build the chip, but their driver track record is terrible. Unless something changes in that department, I don't see them as competition.
 
I highly doubt that would ever happen. You're talking about AMD here, not some small GPU shop.

Nvidia has enough cash on hand to purchase 50% of AMDs stock at the current market rates as of friday.

Nvidia has roughly 1B in cash, friday's AMD market capital at close was 2.13B. Nvidia's market capital at close was 4.88B. Intel on the other hand was 90.13B. AMD IS a small GPU shop.
 
Here's a nice graph to illustrate the need for cache:
cachemem1.gif


You can clearly see that L1 cache accesses are in the order of 1-2 ns. Then you see it hop upwards to about 4 ns for the Core i7, but around 6-8 ns for Yorkfield.
When you hit L3 cache on Core i7, it takes another hit, now around 15 ns... But when you finally hit memory, it's around 60-75 ns.

And that's why even an onboard memory controller needs a big cache. Even L3 cache is still about 4-5 times faster than the onboard triplechannel DDR3.
So I repeat, anyone who says you don't need cache with an onboard controller, is an idiot.
 
Back
Top