AMD Readies FX-8370, FX-8370E Microprocessors.

[TeknoBug]

8 cores? 4.1GHz base clock? 95W? Veeery nice. Might be a worthy upgrade for my Thuban 1045T 95W hex-core CPU. (Already have AM3+ mobo.)

That's not really much of an upgrade, there wasn't that much of a difference when I went from 1090T to 8320. Stick with what you got until AMD releases a new socket (if at all), unless you feel like burning a hole in the pocket with $189.

 

[SOFTengCOMPelec]

http://www.amd.com/en-us/who-we-are/corporate-information/events/amd-30-live
Live stream so you'll have to watch a replay if available
It was during one of the interviews.

Thanks for the link.
I can see the live stream, but can't see any past stuff. But anyway, September 1st is not too far away, so we will know for sure, soon enough.

In my mind (if it is using throttling to get the 95 W TDP figure), the $64,000,000 question is does it throttle, while using normal/standard software, such as gaming or video editing etc ?

If it does throttle (with normal software), then that would make the 95 W part (8370E) most suited to motherboards which can only handle 95 W TDP cpus and/or small PC cases and/or people who want the lowest electricity bills, at all costs.

But if the 95 W TDP part throttles in a well cooled, large PC case (with good quality, powerful VRM motherboard, >=125 W), while using normal software (NOT prime95 etc), I would be rather disappointed. (I mean throttling when using ALL 8 cores, changes to turbo timings would be mostly fine with me).

Of course in the ideal world, the 95 W TDP part, would handle prime95 as well without throttling.

 

[Enigmoid]

IMO, for a desktop CPU you should be able to run an extremely heavy load (prime 95 AVX) and not have the processor use more power than TDP (measured using HW monitor, etc) or throttle below base clock. Otherwise you are mislabeling the CPU.


Mobile is a different story. Intel's 45W quads can run without throttling as can their 35W i5/i3s. Intel U throttles under CPU + GPU load and BT throttles like crazy (like down to 500 mhz CPU, 300 mhz GPU). AMD also throttles quite a bit.

On the desktop throttling is inexcusable. Base clock should be the guaranteed lowest clock under load. If higher clocks under light load are desired lower the base clock and/or lift the boost clock.

 

[SOFTengCOMPelec]

IMO, for a desktop CPU you should be able to run an extremely heavy load (prime 95 AVX) and not have the processor use more power than TDP (measured using HW monitor, etc) or throttle below base clock. Otherwise you are mislabeling the CPU.


Mobile is a different story. Intel's 45W quads can run without throttling as can their 35W i5/i3s. Intel U throttles under CPU + GPU load and BT throttles like crazy (like down to 500 mhz CPU, 300 mhz GPU). AMD also throttles quite a bit.

On the desktop throttling is inexcusable. Base clock should be the guaranteed lowest clock under load. If higher clocks under light load are desired lower the base clock and/or lift the boost clock.

I agree, the TDP should be met, without causing throttling, even with Prime95. Assuming the cpu is adequately cooled and the motherboard can cleanly/reliably supply at least the TDP (ideally more).

What would be good is to have user adjustable TDP (within limits), which I think is coming, in the near future (we may already have such parts available, such as the latest APUs), along with an honest, real life TDP/consumption figure and an upper TDP (Max) limit, based on the cpu package thermal characteristics (and other stuff, such as max junction temperature on the chip die).

i.e. A bit like in Europe (I'm not sure about the US), where stuff like fridge/freezers has labels which show the power consumption on them.

So a cpu would be as follows (in an ideal/better world)..

User Adjustable TDP (throttle) setting, e.g. via bios settings and/or driver software.

Typical (real life) power consumption. E.g. 85 W.

Thermal design power maximum. E.g. 135 W.

Both Intel and AMD (maybe others) are somewhat bad these days, about keeping quiet about the real life power consumption of their cpus. Because the TDP does not necessarily reflect the true/real life figures. (I know it was NOT suppose to anyway, but I want a convenient way of choosing cpus, based on their expected power consumption, and at the moment, the TDP is one of the few ways of doing that, short of googling for reviews or measuring it myself).

Back on topic:
It is good that the market place will see a 95 W, 8 core consumer part, whatever the mechanism is that makes it available (binning, throttling, lower clock frequency, or whatever).

 

[DrMrLordX]

Hmm, that makes the 8320E less interesting as a product as well, if in fact it is differentiated from the standard 8320 by turbo and throttling behavior.

 

[AtenRa]

As i have already said, best product is 8320E, OC to 4.1GHz turbo off and still have 95W TDP at only $140.

 

[DrMrLordX]

If what they're saying is true, the 8320E will be functionally identical to a standard 8320 with turbo/CnQ disabled.

 

[AtenRa]

If what they're saying is true, the 8320E will be functionally identical to a standard 8320 with turbo/CnQ disabled.

Nope, 8320E has 4.0GHz turbo as well. It will not keep turbo at 4GHz the same time interval as 125W TDP FX8320.
It will just throttle down to base quicker than FX8320

 

[ShintaiDK]

As i have already said, best product is 8320E, OC to 4.1GHz turbo off and still have 95W TDP at only $140.

Nope, 8320E has 4.0GHz turbo as well. It will not keep turbo at 4GHz the same time interval as 125W TDP FX8320.
It will just throttle down to base quicker than FX8320

Your statements contradict one another. Not that its any surprise which one is wrong.

 

[Phynaz]

Nope, he is technically correct. Overclock it to 5Ghz and TDP is still 95w. He's using TDP and not power consumption.

 

[ShintaiDK]

Nope, he is technically correct. Overclock it to 5Ghz and TDP is still 95w. He's using TDP and not power consumption.

The TDP is defined at stock clocks. When you OC you also invalidate any TDP because it doesnt apply anymore. So to claim the TDP is still 95W is purely made up out of thin air.

 

[AtenRa]

The TDP is defined at stock clocks. When you OC you also invalidate any TDP because it doesnt apply anymore. So to claim the TDP is still 95W is purely made up out of thin air.

FX8320 = 3.5GHz base, 4.0GHz Turbo 125W TDP

FX8350 = 4.0GHz base, 4.2GHz Turbo 125W TDP

You can OC the FX8320 to FX8350 level and still be at 125W TDP, did you get it now ???

 

[ShintaiDK]

FX8320 = 3.5GHz base, 4.0GHz Turbo 125W TDP

FX8350 = 4.0GHz base, 4.2GHz Turbo 125W TDP

You can OC the FX8320 to FX8350 level and still be at 125W TDP, did you get it now ???

You may and may not be able to depending on the exact chip. You cant just increase the clock of a chip and assume its the same as a higher binned one. There is a reason why AMD bins them as they do.

 

[AtenRa]

You may and may not be able to depending on the exact chip. You cant just increase the clock of a chip and assume its the same as a higher binned one.

Using the same voltage you will have the same TDP and power consumption with both of them.
And i have clearly said to use the FX8320E at 4.1GHz Turbo off specifically for lower voltage that translates to even lower power consumption.

So you can buy the FX8320E at $140, OC to 4.1GHz with Turbo off and have 95W TDP but also lower power consumption than FX8370E. You will just lose a little bit of performance when FX8370 use turbo at 4.3GHz.

 

[ShintaiDK]

Using the same voltage you will have the same TDP and power consumption with both of them.

Thats not correct is it.

 

[AtenRa]

Thats not correct is it.

Using the same voltage and frequency on both CPUs, YES will have the same power consumption and TDP on both of them.

 

[ElFenix]

Using the same voltage and frequency on both CPUs, YES will have the same power consumption and TDP on both of them.

You are seriously abusing the term "TDP"

 

[SOFTengCOMPelec]

Using the same voltage you will have the same TDP and power consumption with both of them.
And i have clearly said to use the FX8320E at 4.1GHz Turbo off specifically for lower voltage that translates to even lower power consumption.

So you can buy the FX8320E at $140, OC to 4.1GHz with Turbo off and have 95W TDP but also lower power consumption than FX8370E. You will just lose a little bit of performance when FX8370 use turbo at 4.3GHz.

Can I check my understanding of what you are saying.

I can buy a £0.25 calculator, remove the (probably) 4 bit cpu. Lower the voltage to 0.0000000000000001 V to minimise its power consumption. Turn its turbo off. Then increase the clock frequency to 5,000,000,000,000,000,000,000,000,000,000 GHz.

Its TDP remains 1 watt (estimate).

So I will have a great super computer for just £0.25, which will last for 1,000,000,000,000,000 million years on a CR2032 battery, and is more powerful than 1,000,000,000,000 Skylakes ?

tl;dr
The 95 w TDP changes, if you start messing from the standard voltage/clock frequency/turbo settings etc, as other(s) have said, above.

 

[AtenRa]

You are seriously abusing the term "TDP"

No i have not, i have clearly distinguish it from Power consumption.

 

[AtenRa]

The 95 w TDP changes, if you start messing from the standard voltage/clock frequency/turbo settings etc, as other(s) have said, above.

Thats why i have said same voltages and no turbo.

 

[Enigmoid]

Using the same voltage and frequency on both CPUs, YES will have the same power consumption and TDP on both of them.

No. There is quite a bit of chip to chip variation. Power consumption will differ.

You also assume that you can use the same voltage on both which may not be possible with a lower binned chip.

 

[AtenRa]

No. There is quite a bit of chip to chip variation. Power consumption will differ.

At same voltage and frequency the difference in power consumption will be negligible, it will also not require a higher TDP rating.

You also assume that you can use the same voltage on both which may not be possible with a lower binned chip.

The only voltage difference will be on the turbo frequency, base frequency will be achieved with almost the same voltage.

That is, the FX8320E will reach 4.1GHz at almost the same voltage as FX8370E.

 

[SOFTengCOMPelec]

Thats why i have said same voltages and no turbo.

Sometimes due to luck (if the 8320 was only fractionally worse than a 8350, and only just missed out on being good enough), or if they have optimized their process so much, that even the lower binned parts, are still very, very good, you may succeed at pretending it was an 8350/8370 all along.
But there are no guarantees, and you may sometimes be unlucky, and not reach the full frequency that the expensive part achieves and/or it uses much more power.

i.e. pay more for a 8350/8370, for guaranteed high clock at reasonable TDP.
Or take your chances on being able to overclock/(underclock and/or undervolt) the cheaper, lower spec part.

Or just live with what the 8320/E is rated for.

 

[Phynaz]

No i have not, i have clearly distinguish it from Power consumption.

Yes, but you are using it as a red herring.

 

[JDG1980]

so you are saying AMD cant design and/or afford to design a new chip?

I wonder how they are paying for all their engineers for the next 2 years to design a brand new uarch but the cant revise their mainstream line with the already designed steamroller cores.

The construction equipment cores are a dead end - an experiment that just didn't work out the way it was hoped. There's no point in throwing good money after bad, especially not for a company as cash-strapped as AMD. It makes a lot more sense to spend what resources they have on K12 R&D than to put it into respinning a poor-selling enthusiast and server implementation of a dying architecture.

K12 and the cat cores are the future. Steamroller/Excavator, at this point, are just stopgaps so that AMD can push the APU concept for the next couple of years until a better architecture is ready.