AMD Unleashes First-Ever 5 GHz Processor

Syphon Filter said:
I need to ask a stupid question which I really should know the answer to, so if the mocking could be kept to a minimum it would save me some blushes...

Question:

How is the Wattage of CPUs calculated and how does it relate to the CPU voltages?

I can't see how a CPU running at circa 1.5V can rack up a Wattage that runs into the hundreds...

P = IV just doesn't stack up as the current figures just get silly (i.e. hundreds of Amps).

Am I an idiot?
Click to expand...

Power (wattage) rises with the square of voltage. Voltage hurts. Power rises roughly linearly with switching frequency. A cpu's turbo algorithm should be designed to stay within the spec'd TDP.

Overly simple math example:

Given: 1.55V @ 5GHz 220W peak tdp full load.
Dropping to 4GHz would be x80% or 176W tdp.
Dropping to 1.31V would be x(1.31/1.55)^2 or 126W tdp.

(Just a rough math example -- not indicative of reality.)
 
It still doesn't make sense to me.

Even at 177Watts that equates to something like 120Amps or thereabouts.

Or is it NOT a P = IV calculation? Which is how DC Wattage is calculated.
 
Ahhhhhh

We're talking THERMAL power. NOT electrical power.

TDP = THERMAL Design Power.

I am an idiot!!
 
Syphon Filter said:
It still doesn't make sense to me.

Even at 177Watts that equates to something like 120Amps or thereabouts.

Or is it NOT a P = IV calculation? Which is how DC Wattage is calculated.
Click to expand...

Right, but I simultaneously increases at the same rate as V, since V = IR, and R is "constant".

Syphon Filter said:
Ahhhhhh

We're talking THERMAL power. NOT electrical power.

TDP = THERMAL Design Power.

I am an idiot!!
Click to expand...

electrical in =~ thermal out
 
Nope, that's not right, if it were we'd all be drawing 100's of amps from our PSUs/at the wall and that patently is NOT the case.
 
So what they are saying is, the processor under full load is capable of outputting up to 220w of heat. Therefore you would need a heatsink capable of removing at least 220w of heat. I'm not a CPU engineer so I can't tell you how much power the CPU draws to create that much heat.
 
Syphon Filter said:
Nope, that's not right, if it were we'd all be drawing 100's of amps from our PSUs/at the wall and that patently is NOT the case.
Click to expand...

It is right. 220W @ 1.5V is 146.7 amps. Or 18.3A @ 12V. Or after 90% efficiency through a PSU, 2.04A from the wall @ 120VAC. 220W of heat is approx 220W power draw.
 
Power is = V * I where V = I *R, so you can also say P = I^2 * R
As the frequency goes up the current draw goes up.
Voltage may be the same at 4Ghz as 5Ghz but the extra current draw at 5 is how the power draw goes up. I don't know of any equation that reliably shows how the current draw will go up with frequency. People have measured power draw and plotted that against frequency but every cpu family will probably have its own characteristics. In other words Phenom II will behave differently than Sandy Bridge which is different than Bulldozer, etc.
 
xorbe said:
It is right. 220W @ 1.5V is 146.7 amps. Or 18.3A @ 12V. Or after 90% efficiency through a PSU, 2.04A from the wall @ 120VAC. 220W of heat is approx 220W power draw.
Click to expand...

This.

Amp x Volt = Watts
 
No, not a "weird troll attempt"

No, sorry, I was serious. I knew AMD still of course made video cards and even have a new architecture coming out. I also know they still supply server grade CPU's and maybe even mainframes and motherboards. I saw something recently in my e-mail that they also have their own branded memory. And the big news that everyone is aware of, they are supplying APU's for the Xbox One and PS4.

I do feel a bit stupid for making my comment but it was done honestly.

I never see AMD around much in regards to consumer CPU's. I also do not see any adverts for new AMD motherboards or things of that nature.

I know a few years ago with the bulldozer release that it was given an absolutely horrid review by everyone. I then saw last many many high ranking AMD people left the company. So when you put all those puzzle pieces together, you can see where many people would just assume they were out of the consumer PC market.

Hopefully this new 5Ghz CPU is as fast as Haswell.
 
If a CPU is running at 3.2ghz and making "X" heat, at 5ghz or 5.2 billion clocks per cycle then current is up 36% and heat production is also up 36% over a CPU at 3.2ghz. Hmm base on that.... We could figure out the power demand of the CPU if we figure what a factory 8350 pulls in wattage of power vs what it produces in heat. So if the 8350 puts out 120w of heat and the 9590 puts out 220. Then the power draw will be..... 80% higher than 8350. Now I just need to lookup what an 8350 pulls in power. Then we could making a sliding scale if voltage is increased.
 
@Syphon Filter: You have the best username ever. I love you. :p

@SixFootDuo:

Nah, AMD never left the market, not anywhere close to that. They did announce a couple years back that they were no longer competing with Intel for the "performance race", i.e. trying to beat Intel's IPC I guess, which makes sense given that they're mainly focusing on APU tech and HSA overall.

AMD has had motherboards made for them on AM3+, FM1, and FM2 sockets (their latest socket). Lots of people still buy AMD CPU's. When Bulldozer (Zambezi) initially released, it received lukewarm reviews due to several major design flaws: single-threaded IPC was lackluster, sometimes being beaten by the Thuban Phenom II x6 chips, high power draw when at full load for the 8150/8120, pipelines were long and unoptimized, branch prediction was bad, and really it wasn't a true upgrade for anyone that already had a decent Phenom II x4 or x6.

However, lots of people who bought the CPU's were evidently pleased regardless, going by the scores on Newegg and Amazon.

Several high-profile engineers have returned to AMD in the recent years. Jim Keller, the best CPU architect ever rejoined the company in 2012 I believe. Raja Koduri, awesome GPU engineer, joined as well. AMD's management is also doing a lot better, especially after they got rid of Dirk Meyer who really started to slip up in his later years of his reign.

I doubt that even this 5GHz Piledriver will do much to knock Haswell away, as it's still the Piledriver architecture. It'll put up some good performance, though, definitely.
 
Sheppard said:
If a CPU is running at 3.2ghz and making "X" heat, at 5ghz or 5.2 billion clocks per cycle then current is up 36% and heat production is also up 36% over a CPU at 3.2ghz. Hmm base on that.... We could figure out the power demand of the CPU if we figure what a factory 8350 pulls in wattage of power vs what it produces in heat. So if the 8350 puts out 120w of heat and the 9590 puts out 220. Then the power draw will be..... 80% higher than 8350. Now I just need to lookup what an 8350 pulls in power. Then we could making a sliding scale if voltage is increased.
Click to expand...

omg just stop, so much wrongness
 
It seems annoying that the most likely people to enjoy and use this would be enthusiasts, yet they want to reserve this for system integrators?
 
rudy said:
It seems annoying that the most likely people to enjoy and use this would be enthusiasts, yet they want to reserve this for system integrators?
Click to expand...

It's not reserved for them, it's only released to them initially. These will be available to everyone else later on.
 
I find this to be very interesting
http://www.bit-tech.net/hardware/2012/11/06/amd-fx-8350-review/7
If you look at the graph, first off at idle the 8350 is the most efficient, full loaded if appears that intel CPUs are pulling quite a bit less until you read the small print at the bottom that states the intel CPUs are tested on Matx boards while the amd were tested on full size atx boards which accounts for 20w difference. Accounting for that the 8350 isn't looking so bad.
Admittedly this has no bearing on what a 5ghz processor looks like on this graph
 
xorbe said:
omg just stop, so much wrongness
Click to expand...

If you can make a post disagreeing with me, then take a few extra seconds and shed light on how you think it is. No conversation is progressed threw disagreement without enlightenment,
 
xorbe said:
It is right. 220W @ 1.5V is 146.7 amps. Or 18.3A @ 12V. Or after 90% efficiency through a PSU, 2.04A from the wall @ 120VAC. 220W of heat is approx 220W power draw.
Click to expand...

The 2A at the wall makes sense. That's a realistic number.

Is the CPU actually drawing 145A @ 1.5V though??? It's that bit that doesn't make sense to me.

@NaroonGTX: Cheers! Had it a long time on various boards! Was one of my favourite games back in PSX days!! Wish they would remake it...
 
Syphon Filter said:
The 2A at the wall makes sense. That's a realistic number.

Is the CPU actually drawing 145A @ 1.5V though??? It's that bit that doesn't make sense to me.
Click to expand...

Yes. 145A is actually within the specifications.

http://en.wikipedia.org/wiki/Socket_AM3

The AM3+ Socket specification contains a few noteworthy design changes over its AM3 predecessor. The 942-pin socket count for the AM3+ is an increase of one over the AM3 Socket's 941-pin socket layout.[9] The AM3+ Socket has larger pin socket diameter of 0.51 mm compared to 0.45 mm with the AM3 Socket. There is a faster serial link of 3400 kHz from the CPU to the power controller, compared to 400 kHz. The AM3+ Socket offers improved power regulation and power quality specifications, including an increased maximum current support of 145 A versus 110 A.
Click to expand...
 
Sheppard said:
If you can make a post disagreeing with me, then take a few extra seconds and shed light on how you think it is. No conversation is progressed threw disagreement without enlightenment,
Click to expand...

You were not talking to me but your post was painful to read to say the least. You don't get certain basic things. Power draw does not go up in a linear relationship to frequency. You are trying to assume it does and you are way, way off. Your assumptions are wrong, plain and simple. Do some research and read some reviews. Here are some examples. Look at the graphs of power consumption vs frequency. These may be old cpus in some reviews, but the the principle has not changed and will not.

http://www.anandtech.com/show/3742/intels-core-i5655k-core-i7875k-overclocked-and-analysed-/3

http://blog.stuffedcow.net/2012/07/ivy-bridge-power-consumption/

http://www.anandtech.com/show/2404/2

http://www.xbitlabs.com/articles/cpu/display/power-consumption-overclocking_10.html
 
http://superuser.com/questions/1635...-increase-as-the-clockrate-of-a-cpu-increases

I read this dudes answer and he seemed to have his stuff fairly straight


pgaster said:
You were not talking to me but your post was painful to read to say the least. You don't get certain basic things. Power draw does not go up in a linear relationship to frequency. You are trying to assume it does and you are way, way off. Your assumptions are wrong, plain and simple. Do some research and read some reviews. Here are some examples. Look at the graphs of power consumption vs frequency. These may be old cpus in some reviews, but the the principle has not changed and will not.

http://www.anandtech.com/show/3742/intels-core-i5655k-core-i7875k-overclocked-and-analysed-/3

http://blog.stuffedcow.net/2012/07/ivy-bridge-power-consumption/

http://www.anandtech.com/show/2404/2

http://www.xbitlabs.com/articles/cpu/display/power-consumption-overclocking_10.html
Click to expand...
 
Sheppard said:
http://superuser.com/questions/1635...-increase-as-the-clockrate-of-a-cpu-increases

I read this dudes answer and he seemed to have his stuff fairly straight
Click to expand...

I did some more research and I was partially wrong and you were mostly right.
This is what I was looking for to start with:

http://www.xbitlabs.com/articles/cpu/display/power-consumption-overclocking_19.html#sect0

"If we look at the power consumption graph during overclocking of almost any processor, we will see, that it consists of two very characteristics parts. The first one is a relatively flat part of the graph that increases slowly, and the second one is the part where the graph starts growing very abruptly. The turning point on this graph doesn’t occur spontaneously, but happens at a very specific moment of time – namely, when it comes to raising the CPU Vcore. In other words, you don’t need t fear that the processor power consumption will skyrocket, until there is real need to adjust any of the system voltages to ensure stability."


So, no increase in core voltage and the power increase is basically linear. Change voltage and power consumption goes up a lot and it's not going to stay linear.

Now back to this new 5Ghz cpu. We don't know the voltage that AMD is running these at. It is going to be the same as the majority of the 8350 cpus? I don't think we know that. Without more information all we can do is speculate really. We just have to wait for AMD to release the specs or until someone can test one of these bad boys.
 
Johan45 said:
That's a proprietary board with 8Gb of fast gddr5. Like I said in an earlier post if AMD can convince the mobo mfgrs to do the same that's the only way they'll be able to continue improving their igpu core. Otherwise the system memory will just stop it in it's tracks.
Click to expand...

Well, sure.

First off, my previous post was mostly in response to the idea that the PS4's GPU will be slower than a 7850, due to memory constrainst. When in fact, it has more bandwidth available than a 7870.

and in response to you: Yes, of course current system memory solutions for laptops and desktop are limiting the upper potential of AMD's APUs. But...so what? For now, they aren't meant to compete with discrete cards. Otherwise, they are by far the best performing "integrated" graphics solutions ever. Which is the point and is pretty darn cool. and I don't see any reason why they wouldn't continue to be, for the foreseeable future. Memory bandwidth aside, we will still see notable improvements with better GPU/APU architectures.
 
Here is where I think my post may have confused you initially, I'm talking specifically about the CPU drawing current based on frequency and voltage. The problem is, to test a processors power consumption, you have to run it on a motherboard, with a heatsink, with other components attached, that being said, we haven't discussed how the added frequency affects the power delivery system of the motherboard and the power consumption there as well, the question then also begs how does phase count, or different power delivery designs in the motherboard also effect power consumption. Another variable is heat and voltage. Increased voltage causes more internal leakage, which is wasted power, heat also contributes to leakage which is power wasted. As the temperature rises the cpu leaks more which causes even more heat and leakage which is even more powere wasted. This is why as we increase voltage it's ever more important to control explicitly the heat variable. This is why cooler chips overclock higher as we know. Therefore the only fair test of CPU power consumption would be to operate each processor with the same power delivery system as well as controlling temperature to as indentical a value as possible. What all those reviews are testing is the power consumption of that specific system with the mother board being a variable, the heatsinks ability to remove heat efficiently being another viariable.
 
Citizen86 said:
I like the Ghz race. 5Ghz turbo chip sounds like great bragging rights :D
Click to expand...

I do not care about GHz. Although I do like to see both AMD and Intel increase the allowable TDP on their top end provided they can keep idle power draw at reasonable levels.
 
You must log in or register to reply here.
Back
Top