Tetrium wrote:

Myth busted...sorry Scali. Maybe you shouldn't accuse people of accusing people of something they didn't do and derailing this otherwise serious technical discussion in the process...again.

Erm, I wasn't attacking anyone in this thread with that remark. And it certainly did not derail anything.
It was just a casual observation. You made a personal attack/insult at me. For which you still have not apologized by the way.

A word of warning about AMD's 'TDP' ratings: They are very arbitrary, and can't be compared directly with Intel (or even other series of AMD CPUs):

As you can see there, under load the system with the 1700X actually draws 137.9W, which is in the ballpark of the 6800k at 138.6W.
Likewise, the 1800X at 155.1W is closest to the 6950X at 156.6W.
AMD however rates their CPUs at 95W TDP, while Intel rates these CPUs at 140W.

If I take this older review of an FX8350, I see that it draws about 182.2W for the entire system: http://www.tomshardware.com/reviews/fx-8350-v … ew,3328-16.html
So Ryzen is less powerhungry than the FX were, but especially with the 1800X, the difference is not that dramatic. They're still pretty powerhungry CPUs (but given the 8 cores, the power-draw is certainly not outlandish).
While the TDP ratings would make you believe that their 8-cores are so power-efficient, that they'd be in the ballpark of Intel's quadcores.

TDP measures have always been somewhat arbitrary, even going back 20 years.

gdjacobs wrote:

TDP measures have always been somewhat arbitrary, even going back 20 years.

'Somewhat', yes, but here we have a difference of 45W in rated TDP, while the measured difference is virtually 0.
It's never been anywhere near this extreme.
Not to mention that 45W on a TDP of 95W is almost 50% off!

F2bnp wrote:

In the next episode of Scali - The Animated Series:

"Who the hell cares about the pricetag? It's nothing more than an arbitrary number, all Intel has to do is lower it!"

🤣

What is the matter with that statement? I know it's an unconventional thought, but it certainly is not illogical. Are you people incapable of taking price out of the equation, and looking purely at a product's technical merits and deficiencies?

Or how do you think technology works?
Price is something you can easily adjust. The chip's architecture is not. You spend years developing, testing and fine-tuning an architecture, and at some point you have to launch a product with it, and it is what it is. You cant change that, unlike price.

So I like to study that architecture.
I mean, do you even get the concept of looking at/being interested in things for what they are, even though you are not interested in buying/owning/using them yourself? Because I most certainly am not interested in buying any new x86 CPU anytime soon (neither Intel nor AMD). But I do like to know what the 'state-of-the-art' is in x86 architectures, and what this means for x86 software I need to develop and optimize in the future.

Intel P4 Prescott 630
TDP - 84W
THG testing:
Min consumption - 86W
Max consumption - 155W
Sysmark Run:
Avg consumption - 119W

http://www.tomshardware.com/reviews/intel-cpu … ion,1750-9.html

gdjacobs wrote:

Intel P4 Prescott 630 TDP - 84W THG Sysmark Run: Min consumption - 86W Avg consumption - 119W Max consumption - 155W […]
Show full quote

Intel P4 Prescott 630
TDP - 84W
THG Sysmark Run:
Min consumption - 86W
Avg consumption - 119W
Max consumption - 155W

What's your point?
TDP is for the CPU only. They measure the entire system here. Obviously the entire system will be higher than just the CPU (motherboard, memory, harddrives, video card etc).
If you compare it to the Core2 Duo however, which is rated at 65W TDP, you see that the total Prescott 630 system draws about 20W more power, exactly what you'd expect on the basis of difference in TDP.

My point was that peak power consumption has absolutely nothing to do with TDP.

gdjacobs wrote:

My point was that peak power consumption has absolutely nothing to do with TDP.

Firstly, did anyone claim otherwise? All I said was that there was a huge discrepancy between the reported TDPs from Intel and AMD. I never said anything about peak.
Secondly, the numbers you quote don't allow you to draw such a conclusion anyway, given that they measure the entire system, over a period of time. Rather than just the CPU, and having fine-grained peak power graphs.
Lastly, it's not very useful to bring in 10-year old data from CPUs that have no bearing to this discussion whatsoever.
Systems have changed a lot over the last decade, also in power consumption/power saving/turbo boosting, cooling etc.

Because if you take TDP literally it means Thermal Dissipation Power (or Thermal Design Power), and is an indication for system builders of how much heat (also measured in Watts) their cooling solution needs to be able to dissipate for the system to remain stable and perform to spec.
It's just that a CPU has no moving parts, so any electrical power applied to its circuits is transferred into heat (because of its internal resistance), meaning that the heat a CPU generates is equivalent to the power it consumes.

Scali wrote:

What is the matter with that statement? I know it's an unconventional thought, but it certainly is not illogical. Are you people […]
Show full quote

What is the matter with that statement? I know it's an unconventional thought, but it certainly is not illogical. Are you people incapable of taking price out of the equation, and looking purely at a product's technical merits and deficiencies?

Or how do you think technology works?
Price is something you can easily adjust. The chip's architecture is not. You spend years developing, testing and fine-tuning an architecture, and at some point you have to launch a product with it, and it is what it is. You cant change that, unlike price.

So I like to study that architecture.
I mean, do you even get the concept of looking at/being interested in things for what they are, even though you are not interested in buying/owning/using them yourself? Because I most certainly am not interested in buying any new x86 CPU anytime soon (neither Intel nor AMD). But I do like to know what the 'state-of-the-art' is in x86 architectures, and what this means for x86 software I need to develop and optimize in the future.

Since you mentioned you hate x86 before, i started to wonder, will you ever ditch your main x86 PC with a non x86 one like a RISC based machine (ARM, IBM POWER... etc)?

Carlos S. M. wrote:

Since you mentioned you hate x86 before, i started to wonder, will you ever ditch your main x86 PC with a non x86 one like a RISC based machine (ARM, IBM POWER... etc)?

Not in the foreseeable future.
I grew up on non-x86 machines, and had to move to DOS/Windows-based x86 machines because all other platforms pretty much died out, and I had to 'blend in' with the rest of the world in terms of what applications and file formats I use.
I don't see the Windows/x86 dominance changing anytime soon, so I don't think a non-x86 machine will be very practical for everyday use.

Scali wrote:

Firstly, did anyone claim otherwise? All I said was that there was a huge discrepancy between the reported TDPs from Intel and A […]
Show full quote

gdjacobs wrote:

My point was that peak power consumption has absolutely nothing to do with TDP.

Firstly, did anyone claim otherwise? All I said was that there was a huge discrepancy between the reported TDPs from Intel and AMD. I never said anything about peak.
Secondly, the numbers you quote don't allow you to draw such a conclusion anyway, given that they measure the entire system, over a period of time. Rather than just the CPU, and having fine-grained peak power graphs.
Lastly, it's not very useful to bring in 10-year old data from CPUs that have no bearing to this discussion whatsoever.
Systems have changed a lot over the last decade, also in power consumption/power saving/turbo boosting, cooling etc.

Because if you take TDP literally it means Thermal Dissipation Power (or Thermal Design Power), and is an indication for system builders of how much heat (also measured in Watts) their cooling solution needs to be able to dissipate for the system to remain stable and perform to spec.
It's just that a CPU has no moving parts, so any electrical power applied to its circuits is transferred into heat (because of its internal resistance), meaning that the heat a CPU generates is equivalent to the power it consumes.

Actually, yes, as you're using instantaneous power measurements to dispute the validity of TDP measurements. Absent of some understanding of the nominal duty cycle anticipated by the TDP measure, this is wrong and has been wrong for a long time.

The benchmark I quoted was just to provide some historical context. The validity of TDP measurements has been a topic of debate for a long time because CPUs (and full systems) have been highly variable in power consumption as a function of load for a long time.

I'm guessing Ryzen includes a throttling capability to allow high peak performance for short durations while still fitting in the specified thermal envelope. Certainly idle power consumption is somewhat more impressive than the comparison samples from Intel (which is actually a significant achievement). Obviously if you want to run SuperPi continuously a stronger thermal management system would be best for peak performance.

I appreciate your technical expertise, but this is complaining about a non issue.

gdjacobs wrote:

Actually, yes, as you're using instantaneous power measurements to dispute the validity of TDP measurements.

Instantaneous? Where did you get that from?
The power consumption was measured at the wall socket, and therefore can in no way be any kind of 'instantaneous' measurement of the CPU (there's too many components in between), but will be some average over a given interval (which I have to admit, they have not specified).
Not to mention that a 'TDP' is not a measurement, it's a specification, and by its nature, somewhat arbitrary to the CPU manufacturer.
As such, there is no 'validity' to speak of. I merely said there's a large discrepancy. That does not imply that the TDP is not 'valid' under the arbitrary constraints that a given CPU manufacturer has chosen.
I'm just pointing out that you shouldn't directly compare ratings of different CPU manufacturers (or even from CPUs of different series from the same manufacturer), and you shouldn't draw conclusions of power draw based on TDP alone. Instead it's better to look at reviewers who use tools to measure actual power draw of the system.
Wouldn't you agree that's good advice? Sounds like you don't, and you keep focusing on the arbitrary TDP.

gdjacobs wrote:

The benchmark I quoted was just to provide some historical context.

It looked more like you yourself don't even know how to read/interpret a power measurement.
You quote the TDP of a CPU, then compare it to system power measurements directly, then draw some wrong conclusions on the data.

gdjacobs wrote:

I appreciate your technical expertise, but this is complaining about a non issue.

Wow, seriously?
I am not talking about small details here, or tight margins. I am talking about a ~50% discrepancy between rated TDP and deltas(!) in measured system power draw.
You can argue all you want, trying to drop fancy buzzwords you googled online to fake some understanding (just as you did with parallelizing physics calculations earlier...). But to me it's obvious that:
1) You don't know what you're talking about.
2) Your only agenda is to try and downplay this issue.

I don't want to get into any details of micro-effects on power consumption, such as 'nominal duty cycle', 'throttling' or whatever, since it's all pretty much irrelevant to the scenario at hand:
A number of systems were tested with recent CPUs (therefore we can assume that they should have 'state of the art' power management, and any differences between products can be seen as shortcomings to meet that state-of-the-art), in the exact same scenario, which was an entire benchmark run of Cinebench R15, which is hardly a detailed 'snapshot' of a few micro-seconds at most.

Bottom line is: the difference is huge. I don't want any 'explanations' of why that difference is huge, or whatever excuses you try to come up with.
The quoted figures are clear, as I said before: you should not expect an AMD CPU rated at a TDP of 95W to be in the ballpark of an Intel CPU with the same TDP. Rather, it is in the ballpark of Intel CPUs rated at 140W TDP.
A 45W discrepancy is not a non-issue, period. End of discussion.
Don't use TDP as substitute for power consumption, as I explained above.

Scali wrote:

What is the matter with that statement? I know it's an unconventional thought, but it certainly is not illogical. Are you people incapable of taking price out of the equation, and looking purely at a product's technical merits and deficiencies?

Are you seriously suggesting we, as buyer, should ignore the price?

Technology doesn't exist in vacuum. Unless CPU's are distributed for free --free as in free beer-- then price will always be consideration, even to those who build computer out of hobbies. Audiophiles have been doing this for decades. Yes, yes, AudioQuest's Diamond speaker cable is probably the best cable around technical-wise, with all the technical merits one can brag about. But I'll be damned if I buy the damn thing, unless I have a pair of JBL Everest to begin with. So no thanks, I'll better spend more money on loudspeakers, because it's what makes the most sonic difference. Same goes with CPU's. Sure, Intel CPU's are cool, but AMD CPU's are cheaper. So I'd rather buy AMD CPU and then add the extra money to my GPU budget. Intel has probably the fastest gaming CPU around, but no thanks, because GPU is what makes the biggest difference.

Isn't TDP Thermal Design Power?
To my understanding, it's the power that the cooler is supposed to dissipate, not the power of the CPU.

It would be interesting to see if this 95W TDP processor is capable of sustaining full load for extended periods of time (let's say 24 or 48 hours) having only 95W TDP rated cooler.

Kreshna Aryaguna Nurzaman wrote:

Are you seriously suggesting we, as buyer, should ignore the price?

No. This is not about buying (read my entire post please), and I don't usually give buyer's advice anyway.

Kreshna Aryaguna Nurzaman wrote:

Sure, Intel CPU's are cool, but AMD CPU's are cheaper. So I'd rather buy AMD CPU and then add the extra money to my GPU budget. Intel has probably the fastest gaming CPU around, but no thanks, because GPU is what makes the biggest difference.

But that's where you're wrong!
My entire point has been that AMD currently only offers 8-core CPUs, which aren't the best option for games.
You want buyer's advice? Listen to this:
The 1700 is their cheapest Ryzen, Newegg currently lists it at $329: https://www.newegg.com/Product/Product.aspx?I … 3-428-_-Product
Now let's pull up some gaming benchmarks: http://www.legitreviews.com/amd-ryzen-7-1800x … eview_191753/10
The Ryzen 1700 gets outpaced by a number of Intel CPUs, even 4-core models, even older ones such as the 2700k.
The 7700k is at the top of these charts, and only marginally more expensive than the Ryzen 1700, at $349: https://www.newegg.com/Product/Product.aspx?I … 7-726-_-Product
So the 7700k is already hands-down the better buy than the 1700X and 1800X: cheaper AND much faster in games.
There are a number of Intel CPUs to choose from below the 7700k, which are cheaper than the Ryzen 1700, and still considerably faster in games.
So if you want to apply your idea of cheaper CPU and spending more on your GPU (which I think makes sense), then AMD is not the one to go for.
Get a cheap Intel quadcore. Heck, even the dualcore i3 7350k isn't doing too badly in these games, and it's only $179.99: https://www.newegg.com/Product/Product.aspx?I … 7-772-_-Product
See?

The 7600k may be the best choice here. Only marginally slower than the 7700k, you still get the 4 cores and 8 threads (and unlocked for overclocking if you like), and it's only $239.99: https://www.newegg.com/Product/Product.aspx?I … 7-728-_-Product

Azarien wrote:

Isn't TDP Thermal Design Power?
To my understanding, it's the power that the cooler is supposed to dissipate, not the power of the CPU.

It would be interesting to see if this 95W TDP processor is capable of sustaining full load for extended periods of time (let's say 24 or 48 hours) having only 95W TDP rated cooler.

Indeed, you are correct. The purpose of TDP is to design CPU cooling system.

Scali wrote:

Instantaneous? Where did you get that from? The power consumption was measured at the wall socket, and therefore can in no way b […]
Show full quote

gdjacobs wrote:

Actually, yes, as you're using instantaneous power measurements to dispute the validity of TDP measurements.

Instantaneous? Where did you get that from?
The power consumption was measured at the wall socket, and therefore can in no way be any kind of 'instantaneous' measurement of the CPU (there's too many components in between), but will be some average over a given interval (which I have to admit, they have not specified).
Not to mention that a 'TDP' is not a measurement, it's a specification, and by its nature, somewhat arbitrary to the CPU manufacturer.
As such, there is no 'validity' to speak of. I merely said there's a large discrepancy. That does not imply that the TDP is not 'valid' under the arbitrary constraints that a given CPU manufacturer has chosen.
I'm just pointing out that you shouldn't directly compare ratings of different CPU manufacturers (or even from CPUs of different series from the same manufacturer), and you shouldn't draw conclusions of power draw based on TDP alone. Instead it's better to look at reviewers who use tools to measure actual power draw of the system.

How can you conclude that there's a discrepancy when there's no valid comparison? You can't compare TDP without an understanding of the caveats involved.

Scali wrote:

Wouldn't you agree that's good advice? Sounds like you don't, and you keep focusing on the arbitrary TDP.

Wow, talk about putting words in my mouth!

Scali wrote:

gdjacobs wrote:

The benchmark I quoted was just to provide some historical context.

It looked more like you yourself don't even know how to read/interpret a power measurement.
You quote the TDP of a CPU, then compare it to system power measurements directly, then draw some wrong conclusions on the data.

I made no conclusions, just quoted the data. As for the data, it's the closest I could find to a similar measure (i.e. bench test power consumption). The only conclusion I made in the following post was that system power load is highly variable depending on CPU load. This is true in both examples.

Scali wrote:

Wow, seriously? I am not talking about small details here, or tight margins. I am talking about a ~50% discrepancy between rated […]
Show full quote

gdjacobs wrote:

I appreciate your technical expertise, but this is complaining about a non issue.

Wow, seriously?
I am not talking about small details here, or tight margins. I am talking about a ~50% discrepancy between rated TDP and deltas(!) in measured system power draw.
You can argue all you want, trying to drop fancy buzzwords you googled online to fake some understanding (just as you did with parallelizing physics calculations earlier...). But to me it's obvious that:
1) You don't know what you're talking about.
2) Your only agenda is to try and downplay this issue.

I don't want to get into any details of micro-effects on power consumption, such as 'nominal duty cycle', 'throttling' or whatever, since it's all pretty much irrelevant to the scenario at hand:
A number of systems were tested with recent CPUs (therefore we can assume that they should have 'state of the art' power management, and any differences between products can be seen as shortcomings to meet that state-of-the-art), in the exact same scenario, which was an entire benchmark run of Cinebench R15, which is hardly a detailed 'snapshot' of a few micro-seconds at most.

Bottom line is: the difference is huge. I don't want any 'explanations' of why that difference is huge, or whatever excuses you try to come up with.
The quoted figures are clear, as I said before: you should not expect an AMD CPU rated at a TDP of 95W to be in the ballpark of an Intel CPU with the same TDP. Rather, it is in the ballpark of Intel CPUs rated at 140W TDP.
A 45W discrepancy is not a non-issue, period. End of discussion.
Don't use TDP as substitute for power consumption, as I explained above.

I'm not going to get into an argument over qualifications. Granted, you're much more knowledgeable in game programming, but my understanding of physics and electrical engineering is undoubtedly stronger than yours. You can choose to believe that or not.

I contend that you cannot evaluate the power consumption and thermal load of these CPUs absent of an understanding of how they will be operated. Cinebench is not necessarily indicative of a typical load for anyone but Cinebench users (who are actively running benchmark jobs 100% of the time). It MIGHT be somewhat reflective of users running jobs in Cinema 4D continuously.

The reality is that some users will be interested in lower idle power consumption and some will be more concerned with full load consumption. Most people will be somewhere in between. TDP is at most a valid comparison between CPUs in the same family.

As for agenda, I have none. I did speculate on the potential basis for TDP assignment. I'm not going to comment on YOUR agenda. Your ACTIONS have been to over react based on a comparison between two company's arbitrary numbers and to cherry pick data in general that supports your preference for Intel hardware.

Scali wrote:

No. This is not about buying (read my entire post please), and I don't usually give buyer's advice anyway. […]
Show full quote

Kreshna Aryaguna Nurzaman wrote:

Are you seriously suggesting we, as buyer, should ignore the price?

No. This is not about buying (read my entire post please), and I don't usually give buyer's advice anyway.

Kreshna Aryaguna Nurzaman wrote:

Sure, Intel CPU's are cool, but AMD CPU's are cheaper. So I'd rather buy AMD CPU and then add the extra money to my GPU budget. Intel has probably the fastest gaming CPU around, but no thanks, because GPU is what makes the biggest difference.

But that's where you're wrong!
My entire point has been that AMD currently only offers 8-core CPUs, which aren't the best option for games.
You want buyer's advice? Listen to this:
The 1700 is their cheapest Ryzen, Newegg currently lists it at $329: https://www.newegg.com/Product/Product.aspx?I … 3-428-_-Product
Now let's pull up some gaming benchmarks: http://www.legitreviews.com/amd-ryzen-7-1800x … eview_191753/10
The Ryzen 1700 gets outpaced by a number of Intel CPUs, even 4-core models, even older ones such as the 2700k.
The 7700k is at the top of these charts, and only marginally more expensive than the Ryzen 1700, at $349: https://www.newegg.com/Product/Product.aspx?I … 7-726-_-Product
So the 7700k is already hands-down the better buy than the 1700X and 1800X: cheaper AND much faster in games.
There are a number of Intel CPUs to choose from below the 7700k, which are cheaper than the Ryzen 1700, and still considerably faster in games.
So if you want to apply your idea of cheaper CPU and spending more on your GPU (which I think makes sense), then AMD is not the one to go for.
Get a cheap Intel quadcore. Heck, even the dualcore i3 7350k isn't doing too badly in these games, and it's only $179.99: https://www.newegg.com/Product/Product.aspx?I … 7-772-_-Product
See?

Just as most computer users in the real world out there, my rig is not solely for gaming. I probably spend more time doing office works than gaming. On the other hand, Tech Report has concluded that Ryzen has better price/performance ratio for non-gaming applications. So nope, I'm still not buying Intel. Of course, nobody said you cannot go to the Tech Report's website and contradict their conclusion. Believe me, I'll be very interested in the outcome.