AMD unleashes first ever commercial “5GHz” CPU, the FX-9590

[grimpr]

2 issues.

As said above, you confuse power and temperature. Basic thermodynamics.

And AMD measures their temperature a different place and adds an offset. So the 2 cant be compared. Not to mention AMD is unwilling to define any specs for the same reason.

I dont confuse nothing. You are confused about reality. Compare average and 100% load temperatures on FX & Haswell and then come back.

 

[ShintaiDK]

I dont confuse nothing. You are confused about reality. Compare average and 100% load temperatures on FX & Haswell and then come back.

A 2000C needle will not heat your room much. A 50C large radiator will.

The heat output on FX CPUs are almost twice as high. Hence they are the space heater in the room.

 

[grimpr]

A 2000C needle will not heat your room much. A 50C large radiator will.

The heat output on FX CPUs are almost twice as high. Hence they are the space heater in the room.

Well, big mouth small talk, prove it.

This is a tech forum. Stick to technical arguments and leave the tough guy act at home
-ViRGE

 

[ShintaiDK]

Well, big mouth small talk, prove it.

IDC already proved it. Just lookup his posts. Or MSIs support response to overheating VRMs.

 

[ClockHound]

Space heaters are good. Just checked with NASA. Space is very, very cold. An orbital FX could be the new rig for me. Screw cloud computing, I want space computing. But with sub-space latency.

 

[AtenRa]

IDC already proved it. Just lookup his posts. Or MSIs support response to overheating VRMs.

Actually IDC has only proved that the PSU will draw more than 125W from the wall, not the CPU.

 

[Meghan54]

Actually IDC has only proved that the PSU will draw more than 125W from the wall, not the CPU.

So, guess IDC just showed the motherboard and memory use a lot of power, eh? Or maybe he just had a crappy psu that was only around 60% efficient?

 

[ViRGE]

I just gave this thread a good, deep clensing. Take it down a notch folks and drop the hostility, otherwise I'll have to hand out more vacations

-ViRGE

 

[gevorg]

OK, I'll ask a noobish question:

Instead of raising the top clock rate so high, why can't AMD increase the die size and add moar cores/modules? Wouldn't that be more power/performance efficient than a 220W+ TDP chip? Sure, the price will be higher, but its a premium product anyway.

 

[itsmydamnation]

The heat output on FX CPUs are almost twice as high. Hence they are the space heater in the room.

"Heat" is not linear. producing twice as much energy as heat does not produce twice as much heat. At least get your basic year 7 physics right.

OK, I'll ask a noobish question:

Instead of raising the top clock rate so high, why can't AMD increase the die size and add moar cores/modules? Wouldn't that be more power/performance efficient than a 220W+ TDP chip? Sure, the price will be higher, but its a premium product anyway.

if this chip has any changes at all its likely just to be in metal layers. what your talking requires significantly more resources/time. we will have to wait until next year to see what steamroller "server" SOC looks like but i wouldn't be surprised if it was 5-6 modules, at the same time i wouldn't be surprised with 4 either.

 

[ViRGE]

"Heat" is not linear. producing twice as much energy as heat does not produce twice as much heat. At least get your basic year 7 physics right.

Semiconductors are for all practical purposes perfect resistors when it comes to power and heat. They do no useful mechanical work, so virtually all power they consume is converted to heat in the process of doing their job. So a processor that consumes twice as much power would generate twice as much heat (not to be confused with temperatures).

Wikipedia on resistors: "The total amount of heat energy released over a period of time can be determined from the integral of the power over that period of time:"

 

[Abwx]

A lot of people by there are assuming that this chip
is some kind of overclocked 8350 , a 125W chip that
would be pushed to 5Ghz thanks to overvolting.

That wont be the case , the principle will be to use
chips that work at 8350 frequencies but with 10%
lower voltage , this will reduce TDP by 21% down
to 100W , then pushing frequency by 20% will bring
TDP back to 120W , assuming 10% extra voltage
is required for full stability this will increase TDP
by 21% to 145W , far from the estimations that
can be seen here and there and wich are based
on mythicals interpretations of physics laws.

 

[zebrax2]

OK, I'll ask a noobish question:

Instead of raising the top clock rate so high, why can't AMD increase the die size and add moar cores/modules? Wouldn't that be more power/performance efficient than a 220W+ TDP chip? Sure, the price will be higher, but its a premium product anyway.

Adding more cores means more R&D cost, money that AMD has very little right now. Also AMDs performance problem isn't that they don't have enough core its that they are weaker in comparison with the competion and jacking up the clocks helps aleviate that

 

[itsmydamnation]

Semiconductors are for all practical purposes perfect resistors when it comes to power and heat. They do no useful mechanical work, so virtually all power they consume is converted to heat in the process of doing their job. So a processor that consumes twice as much power would generate twice as much heat (not to be confused with temperatures).

Wikipedia on resistors: "The total amount of heat energy released over a period of time can be determined from the integral of the power over that period of time:"

i know how a MOSFET works, I have never seen someone use the word heat when talking about the measuring of transfer of energy, always joules, watts etc . i suppose it comes down to what you define "heat" as. The point is producing twice as much energy doesn't not produce twice as much "heat" ( to me heat is the end result of energy transfer).

 

[ViRGE]

i know how a MOSFET works, I have never seen someone use the word heat when talking about the measuring of transfer of energy, always joules, watts etc . i suppose it comes down to what you define "heat" as. The point is producing twice as much energy doesn't not produce twice as much "heat" ( to me heat is the end result of energy transfer).

Technically speaking you are correct. Temperature is the amount of thermal energy within an entity, while heat is the transference of it.

However every watt (or rather joule) of power utilized by a CPU is going to be converted to thermal energy. And it's a continuous process, so you don't get to store that thermal energy anywhere. It has to leave the CPU and the heatsink. So regardless of poor naming, you still have to send that thermal energy somewhere, and hence in common parlance heat. In any case, for computer energy consumption and cooling this is considered to be a reasonable approximation/dumbing down of how thermodynamics works.

 

[galego]

Haswell at 4.0 would match this while using less than half the power. no one cares about the 3960x as even an older 2600k can basically match the 3960x overall in games when both the cpus are oced.

Haswell overclocked at 4.0 would match the centurion at stock in some tasks but not in others. It would use about twice the power if you ignore the rest of the hardware. Add a single high-end graphic card (e.g., that mentioned in my previous post) and the total power consumption (ignoring screen) would be only about a 20% more.

It is untrue that no one cares about the 3960x.

"Overall in games" because the old games are poorly threaded and ignore the extra cores/threads. But gaming design is changing and next gen is heavily multithreaded.

 

[Maximilian]

Haswell at 4.0 would match this while using less than half the power. no one cares about the 3960x as even an older 2600k can basically match the 3960x overall in games when both the cpus are oced.

*Looks down at hexcore rig*

I care :'(

 

[galego]

IDC already proved it. Just lookup his posts. Or MSIs support response to overheating VRMs.

IDC proved nothing and the supposed MSI issue was already debunked in the forums: a supposed email written by someone at 'support', confounding TDP with power consumption, sent to a supposed buyer with a failing MSI mobo.

 

[Name User]

*Looks down at hexcore rig*

I care :'(

dude you better get cinema 4d out and start rendering bowling balls being dropped into paint filled bathtubs otherwise you'll never make sense of that purchase

 

[guskline]

IDC validly tested his own FX8350 for power draw. To say he proved nothing is FALSE!

 

[ShintaiDK]

IDC proved nothing and the supposed MSI issue was already debunked in the forums: a supposed email written by someone at 'support', confounding TDP with power consumption, sent to a supposed buyer with a failing MSI mobo.

I have never seen it debunked. However I have seen edvidence that AMD sells out of spec CPUs. And that AMD refuses to publish specs. Most likely to try and hide the issues and hope the consumer dont realise it and blame someone else instead (Like mobo makers.). But with a 220W 4.7Ghz base clock CPU. The desperation is deep in that company.

 

[AtenRa]

However I have seen edvidence that AMD sells out of spec CPUs.

What evidence is that ??? IDCs entire system power usage ??? Or MSI's motherboard inability to use a 125W CPU ??

Stop the FUD.

 

[AtenRa]

So, guess IDC just showed the motherboard and memory use a lot of power, eh?

Can you show us how much power the motherboard and the memory used out of that total power drawn by the PSU for the entire system ??

 

[nemesismk2]

I can't wait to see a review of the AMD FX 9590 at Anandtech

 

[SlitheryDee]

I seem to remember there being rumors of 5g P4 CPU's floating around somewhere in europe. I remember it specifically because the article said they were probably "monstrosities" (not in a good way) even if they really existed. I guess this makes AMD king of the monsters if nothing else.