AMD Piledriver rumours ... and expert conjecture

Page 168 - Seeking answers? Join the Tom's Hardware community: where nearly two million members share solutions and discuss the latest tech.
Status
Not open for further replies.
We have had several requests for a sticky on AMD's yet to be released Piledriver architecture ... so here it is.

I want to make a few things clear though.

Post a question relevant to the topic, or information about the topic, or it will be deleted.

Post any negative personal comments about another user ... and they will be deleted.

Post flame baiting comments about the blue, red and green team and they will be deleted.

Enjoy ...
 
Read the interview again. He wasn't talking about a dGPU but about using the iGPU for these things. The iGPU shares the exact same memory as the CPU, same bus and access paths. Currently the iGPU use's a different virtual address space for it's memory and copying data between the two requires page locks and framing. By allow the GPU and CPU to talk in the same address space you can treat data in system memory as data in GPU memory and vice versa. This is something that will be implemented at the kernel level first and foremost. You could also use this within the 2GB limit of 32-bit land, but would require some form of address translation similar to what PAE does.

Honestly, we all need to get the hell away from 32-bit world, 2GB is not enough anymore.
Whats nice here is this also brings many kernel initiations
 

amd trinity use turbo, means too much variable frequency

so it doesn't looks like an overclock fail

but if it does, then we will need unlocked (variable) capacitor and inductors.

And lc circuit oposses any change in phase/direction of current so there will be no cnq, no turbo either, if it interfare with cpu clock frequency
 
Any OC results out there for Trinity?

1. I can't put the smiley face at the top of pg 100 without deleting peoples happy comments ... thats not fair.

2. I'll have had an aneurism well before pg 200 ...

3. We will close this one as soon as PD cores show up and some benchies are put up.

4. Then one of you can start a new Sticky ... maybe one for Mob and one for Desktop ... by then the new battle lines will be well drawn.

:)
 
Which is why I am happy with my HD7970 even though the GTX680 is a bit better. Well at "normal" resolutions but at super high, the extra 1GB of VRAM helps a bit.

So strangley I currently have a brand loyalty to AMD GPUs. And Corsair stuff. Corsair makes great stuff.

I will go along with this, loyalty or preferences;

CPU: Well I can afford multi setups so I have Intel and AMD, but there is a romance or irrational love towards the AMD chips.
MOBO: Asooos or MSI
RAM: G.Skill, Patriot
PSU: Antec or XFX(cheaper than seasonic with by and large the same output)
GPU: Was Nvidia, now AMD sold on the Sapphire HD7950 Toxic OC Edition, its a absolute biscuit.
Chassis: Silverstone (just so dang expensive TJ07/11 greatest case ever made, unless you go custom)
HD: doesn't matter
SSD: Plextor, Intel(I feel dirty), Crucial, Patriot.

Nvidia/AMD

I had a conversation with a long time reviewer, one of the best overclockers around and the issue was on Nvidia vs AMD sales, essentially Nvidia make more on Graphics cards than what AMD do, this was the premise for the slowness of Kepler production. I don't fully believe that, I think there is some smoke around the degredation issue and FABS supply shortages. I have posted this elsewhere, but have had two GTX 680 reference design cards fail at around 1400mhz on the core, sure it is probably batch related but something smells funky, and it ain't last nights dinner.



 

desismileys_3838.gif
 
Just going on some of the threads in this section, I can only see AMD making a legitimate counter challenge to Intel, from 2014, by that stage AMD should have;

a) Management structure.
b) Implementation and Planning
c) Stable resources

Right now it is all hands on deck, but it is not like they are going quietly like some people are suggesting, its the taking stock phase.
 
Looks like a basic circuit. Cannot say. Numbers are figures, however!

Cannot say until the silicon is out for testing. BTW, how can one calculate a frequency of such device. I know basic sine functions, but how does this device increase/decrease frequency?

An LC tank circuit doesn't generate oscillations on its own.....it rather 'favours' one selected frequency, called the resonant frequency, and retards all others.

Basically, if u take an signal generator and apply a signal, say +5v to an LC circuit, and measure the output, u'll observe this...

As your increase the applied freq from low to high, the output voltage gradually begins to rise(say from 1volt upwards).

At the resonant freq, there will be a sudden and sharp rise in output voltage (say a peak at 4.5volts)

After the resonant freq, the output voltage drop sharply first, and then will gradually drop to zero.

just ask an electronic engineer, if u happen to have one in ur friends circle, he'll be happy to explain:)
 
amd trinity use turbo, means too much variable frequency

so it doesn't looks like an overclock fail

but if it does, then we will need unlocked (variable) capacitor and inductors.

And lc circuit oposses any change in phase/direction of current so there will be no cnq, no turbo either, if it interfare with cpu clock frequency


your missing the point, there's a reason why its called turbo!!! the cpu can't run at that speed constantly, it will overheat!!! Its only at elevated freq's for shorts durations,and quickly comes back to stock once power/temps go beyond allowable limits.

It does "interfere", it just will draw more power at other frquencies than at the resonant freq. :)

PS:- I highly doubt one can implement smoothly variable caps and inductors in silicon!! maybe 3 or 4 diff cap values may be present, and any one can be selected depending on the freq, by using some form of analog mux ckt.
 
An LC tank circuit doesn't generate oscillations on its own.....it rather 'favours' one selected frequency, called the resonant frequency, and retards all others.

Basically, if u take an signal generator and apply a signal, say +5v to an LC circuit, and measure the output, u'll observe this...

As your increase the applied freq from low to high, the output voltage gradually begins to rise(say from 1volt upwards).

At the resonant freq, there will be a sudden and sharp rise in output voltage (say a peak at 4.5volts)

After the resonant freq, the output voltage drop sharply first, and then will gradually drop to zero.

just ask an electronic engineer, if u happen to have one in ur friends circle, he'll be happy to explain:)
Planing on being a future electrical engineer...
:??: Not so smart(me) I guess...

 
amd trinity use turbo, means too much variable frequency

so it doesn't looks like an overclock fail

but if it does, then we will need unlocked (variable) capacitor and inductors.

And lc circuit oposses any change in phase/direction of current so there will be no cnq, no turbo either, if it interfare with cpu clock frequency

Planing on being a future electrical engineer...
:??: Not so smart(me) I guess...

The nature of a capacitor is to offer very low resistance at the begin of the phase of oscillation, an inductor offers high resistance to any change of current which is the case at the begin of the phase of oscillation. As the oscillation phase progresses, the roles reverse. Thus, the capacitance and the inductance counter each other. This works best at one frequency which is called the resonant frequency and depends upon the level of capacitance and inductance. This is very basic electronics and can be seen with an oscilloscope.

I don't know how the inductance and capacitance has been manufactured into the cpu structure but this is the principle used.

 
I don't know how the inductance and capacitance has been manufactured into the cpu structure but this is the principle used.

i had done a small course on VLSI design and manufacturing, about a year ago :)

Capacitance are made either by sandwiching 2 sufficiently large metal layers, with an insulating layer in between. Another way to create smaller capacitance is to use 2 thin and long finger like metal areas on the same layer.

Inductances on the other hand, are formed my drawing a square shaped 'spiral' coil on the same metal layer:)

 
I don't know how the inductance and capacitance has been manufactured into the cpu structure but this is the principle used.

i had done a small course on VLSI design and manufacturing, about a year ago :)

Capacitance are made either by sandwiching 2 sufficiently large metal layers, with an insulating layer in between. Another way to create smaller capacitance is to use 2 thin and long finger like metal areas on the same layer.

Inductances on the other hand, are formed by drawing a square shaped 'spiral' coil on the same metal layer:)
 

Oh man, that looks hot...

It even has a lil' steampunk aura around it, hahaha.

And regarding the AMD (not)news that came around in a thunder blaze, I'm sure they won't stop developing better CPUs, but they'll include the consumption and graphics metrics in the mix. That's actually a good thing for us and Intel (telling you all, again) recognized from the APU road AMD took a while ago with SB and IB and now will continue with Haswell.

Like the non-biased article here at Toms points out (thanks for that), ARM is also lurking the waters, emerging as a low power and low cost alternative on each generation to X86. The Software (OS+Apps) platform that surrounds it is very healthy and growing fast. So healthy in fact that MS and Intel are worried about it and have started a race (Win8 and ULV Atoms) to not get caught with their guards down. Well, MS got it the hard way, but Intel stepped up in time to the game (go take a look at the Lenovo Xolo). AMD was in the game, but gave the cookie to Qualcomm when they got ATI. At least they're in very good terms with ARM, so I'm sure the times ahead, if Mr. Read plays his cards nicely, will be very good for the mobile market from AMD's side.

Intel will dominate the "Desktop" market until the "Mobile" (Low power -> portables and enterprise mostly) becomes the most important segment and replaces the Desktop as we know it.

Cheers!
 
Oh man, that looks hot...

It even has a lil' steampunk aura around it, hahaha.

And regarding the AMD (not)news that came around in a thunder blaze, I'm sure they won't stop developing better CPUs, but they'll include the consumption and graphics metrics in the mix. That's actually a good thing for us and Intel (telling you all, again) recognized from the APU road AMD took a while ago with SB and IB and now will continue with Haswell.

Like the non-biased article here at Toms points out (thanks for that), ARM is also lurking the waters, emerging as a low power and low cost alternative on each generation to X86. The Software (OS+Apps) platform that surrounds it is very healthy and growing fast. So healthy in fact that MS and Intel are worried about it and have started a race (Win8 and ULV Atoms) to not get caught with their guards down. Well, MS got it the hard way, but Intel stepped up in time to the game (go take a look at the Lenovo Xolo). AMD was in the game, but gave the cookie to Qualcomm when they got ATI. At least they're in very good terms with ARM, so I'm sure the times ahead, if Mr. Read plays his cards nicely, will be very good for the mobile market from AMD's side.

Intel will dominate the "Desktop" market until the "Mobile" (Low power -> portables and enterprise mostly) becomes the most important segment and replaces the Desktop as we know it.

Cheers!

/sigh Chuck Norris > Me, I will admit.

And yeah people been taking that news completely the wrong way. CPU's do have more then enough processing power today, software just hasn't been able to use that power. Take "gaming" for example, gaming traditionally push's the envelope in consumer desktop processing power. An Intel SB i5-2500K has four cores, yet most games will only use one to two cores. Thus 50% of that i5's processing power is unused. It gets even worse on something as wide as a BD uarch, games were using 25% of that chips processing power leaving the rest underutilized. The results of the processor arms race is that we now have these extremely powerful processors that very few software programs can even fully utilize. Console's didn't help that issue at all.

What AMD is doing now (and Intel is following suite) is focusing more on lower power content consumption computer. Mobile / laptop type devices that let you do everything a desktop can only in a much more comfortable form factor. Instead of six or eight cores with 25~50% of the power being unused, we're going to go with three to four and use the remaining space for a HW vector coprocessor for accelerating content consumption. And while the desktop market will still exist, especially for high end gaming, its more of a niche market now. Gaming has been about GPU's more then CPU's for years now, people talking about $500~$700 USD GPU's and $200 USD CPUs kinda indicates that.
 
I've never measured VRAM usage to be honest, but with iGPUs it could be an important thing to take into account in the future.

Cheers!
And until some possible ram on die or dedicated somewhere on the Mobo, your left with using the system ram. AMD's supported memories speeds are higher, but their memory controller could really use some work, especially with all the focus on APU's. It would be very beneficial to do some work on it in the near future.
 
We have had several requests for a sticky on AMD's yet to be released Piledriver architecture ... so here it is.

I want to make a few things clear though.

Post a question relevant to the topic, or information about the topic, or it will be deleted.

Post any negative personal comments about another user ... and they will be deleted.

Post flame baiting comments about the blue, red and green team and they will be deleted.

Enjoy ...

who the green team?
 
Status
Not open for further replies.