Production of AMD "Piledriver" FX Processors Begin Q3 2012

[Fourstaff]

Dude read my post again. As far as Phenom II and Bulldozer you just agreed with me



Also, if we are in agreement the 9650, Phenom II and Bulldozer perform about the same (atleast in single threaded games) That means it's impossible for Phenom II to be 25% faster than Bulldozer as Hustler suggested

That is what I get when I skim through posts :shadedshu

Well, there will always be fanboys everywhere, and aggressive weeding is needed every so often.

Either way, AMD needs to boost their singlethreaded heavily to compete with Sandy Bridge past the $150 dollar mark, but I think AMD has given up in that segment and focus their resources in server and also low to mid end. Bulldozer while lacking in high end gaming does very well in mid and below, trading blows with the i3 and non-k i5 when overclocked. Power consumption still needs to be improved by 50% to make it truly competitive, no just on benches.

 

[Dent1]

First of all i said 'up to' 15% slower, meaning in some CPU bound benchmarks, 1 Bulldozer core is 15% slower than 1 Phenom II core.

The benchmarks are out there if you look....

Fair enough. But saying "upto" is misleading. Because upto isnt on average.

Lets say we bench 20 games. Phenom II wins only 5 tests by a 15% lead, and Bulldozer wins 15 test by a 10% lead. Would it be fair to say Phenom II is upto 15% faster when on average it got spanked?


Fourstaff, no offense taken. With all the fanboys around sometimes we automatically go on the defensive. I've been guilty of it myself.

I agree singlethreaded performance needs to be improved heavily. Trinity is in the right step, as I pointed out earlier, Anandtech suggests a 20-25% gap between Trinity and Ivy in CPU. So hopefully AMD use this time to perfect the Piledriver between now and then to close the gap.

 

[babash*t]

I reserve my opinion until I see the Piledriver FX chips. Glad to see they are going to stick with AM3+ for at least another generation.

Problem with sticking with am3+ is no quad channel memory

 

[librin.so.1]

Problem with sticking with am3+ is no quad channel memory

On the bright side - mobos currently housing our bulldozers won't be obsoleted for a longer time.
[warning! Some philosophical thoughts ahead!]
I know that this has some bad implications, but let us not forget to always also look for the bright side. There is always two ends to a stick.
That means "good" things also carry some "bad" implication. So, let's just be thankful for what we have - as in the first place, we have something ;]

 

[Dent1]

Problem with sticking with am3+ is no quad channel memory

Bingo.

Apparently the server chips will support quad channel. I'd be interested to find out how the dual channel desktop version performs in comparison to the quad channel server variant.


http://www.techpowerup.com/159062/A...nnel-DDR3-IMC-G34-En-Route-Desktop-.html?cp=3

 

[pantherx12]

My predictions !

4 ghz stock speed ( top end model)
+20% IPC

5GHZ on air ( I can nearly do that already)

Still will run as hot as BS but not AS hot as BS.

 

[librin.so.1]

Still will run as hot as BS but not AS hot as BS.

I don't get it.

 

[Hustler]

Lets say we bench 20 games. Phenom II wins only 5 tests by a 15% lead, and Bulldozer wins 15 test by a 10% lead. Would it be fair to say Phenom II is upto 15% faster when on average it got spanked?

Well I'm not really bothered about modern gaming anyway, the GPU is the biggest factor there these days, and when it isn't, your probably talking about the difference between 30fps and 33fps.

I need raw single thread CPU performance because i heavily use emulators, which are usually not very multi core efficient, so IPC is by far the most important criteria for my needs.

As much as i would love Trinity to match a 2600k on a clock for clock IPC, there is just too big a gap for that to happen by Q4 2012.

 

[babash*t]

On the bright side - mobos currently housing our bulldozers won't be obsoleted for a longer time.
[warning! Some philosophical thoughts ahead!]
I know that this has some bad implications, but let us not forget to always also look for the bright side. There is always two ends to a stick.
That means "good" things also carry some "bad" implication. So, let's just be thankful for what we have - as in the first place, we have something ;]

True, true

 

[Dent1]

As much as i would love Trinity to match a 2600k on a clock for clock IPC, there is just too big a gap for that to happen by Q4 2012.

Bear in mind Anandtech's assessment was based Ivy Bridge vs Trinity (20-25% CPU difference). The gap between Sandy Bridge and Trinity would be smaller maybe 15-20% taking a conservative guess. So it's not impossible to think a desktop Piledriver could be 10-15% faster than Trinity. Thus atleast catching upto Sandy Bridge (2600k) or on par. But yet being behind Ivy Bridge somewhat.

 

[babash*t]

Bingo.

Apparently the server chips will support quad channel. I'd be interested to find out how the dual channel desktop version performs in comparison to the quad channel server variant.


http://www.techpowerup.com/159062/A...nnel-DDR3-IMC-G34-En-Route-Desktop-.html?cp=3

Well if they confirm the 10core vishera for desktop, and have it out and available by Q3/4 I don't think lack of quad channel will be a problem for me

 

[librin.so.1]

Well if they confirm the 10core vishera for desktop,[...] I don't think lack of quad channel will be a problem for me

10 core would mean 10mb L2 cache. I wonder, if the L3 cache would still be 8mb on a 10core/5module die. If handled well, a large L2 & L3 can reduce the amount of situations when 2-chanel memory would be a bottleneck, I think.
That said, I have no idea if bulldozer handles its L2 & L3 caches well. Along with not remembering if cache handling improvements were mentioned in piledriver's improvement highlights. xD

 

[pantherx12]

I don't get it.

The processor runs incredibly hot when over clocked.

I expect the new processor will run hot too.

But not as hot.

 

[librin.so.1]

Ah, thanks for clearing that ambiguity of Your previous post up. ^^^^

 

[cadaveca]

the implementation of resonant clock mesh technology

Hrm...curious....tell me more, tell me more.

 

[librin.so.1]

[...] the implementation of resonant clock mesh technology[...]

Hrm...curious....tell me more, tell me more.

Each time I hear about something new that sounds like it will be fiddling with the clock speeds behind the user's back, I frown.
Many such things already made overclocking more and more confusing with each such additional technology. If this trend keeps up, in a few years time, overclocking can turn into "I don't even f-ing know if those changes I made affect anything at all". One such "bad" example of a chip getting near (bot not quite there yet) this territory is Nv Kepler.

It will be a sad day, when overclocking becomes a thing comparable to "voodoo magic done by an amateur using a 'cookbook'.". :shadedshu

 

[pantherx12]

Each time I hear about something new that sounds like it will be fiddling with the clock speeds behind the user's back, I frown.

That isn't what it does.

"Electricity passes from the capacitor to the inductor, where it creates a magnetic field. When the capacitor’s charge reaches zero, the current flow reverses — power shifts from the inductor to the capacitor, and the magnetic field dissipates. This process is analogous to the movement of a pendulum or the sloshing of water in a tank, which is where the “tank circuit” moniker comes from.

Here’s the key: It takes significantly less power to keep the pendulum swinging than it does to start over from a neutral position each and every cycle. Cyclos’ design reportedly cuts clock distribution power by “up to 24% while maintaining the low clock-skew target required by high-performance processors.” Cyclos claims that using its technology can cut total IC power by up to 10%. "


Basically a power saving measure.

Saved power also means lower temperatures.


All goes towards them being able to clock BD higher.

 

[eidairaman1]

for a lil easier explaination the inductor is a wire coil wrapped around a core core (the donuts on motherboards are a good example, Magnetic fields are called flux fields (Coils are used to step up or step down voltage/current in Transformers)

That isn't what it does.

"Electricity passes from the capacitor to the inductor, where it creates a magnetic field. When the capacitor’s charge reaches zero, the current flow reverses — power shifts from the inductor to the capacitor, and the magnetic field dissipates. This process is analogous to the movement of a pendulum or the sloshing of water in a tank, which is where the “tank circuit” moniker comes from.

Here’s the key: It takes significantly less power to keep the pendulum swinging than it does to start over from a neutral position each and every cycle. Cyclos’ design reportedly cuts clock distribution power by “up to 24% while maintaining the low clock-skew target required by high-performance processors.” Cyclos claims that using its technology can cut total IC power by up to 10%. "


Basically a power saving measure.

Saved power also means lower temperatures.


All goes towards them being able to clock BD higher.

 

[Widjaja]

Can only hope AMD don't make Piledriver the chip which Bulldozer should have been which is a step up from Phenom II

 

[Syborfical]

In the old reviews where they did benchmarks under linux, bulldozer did really well. There it did beat Nehalem & Sandy Bridge on numerous tests, and was head to head in most where it did not beat 'em, losing in just a few. It did especially well in compiling. That is, IIRC.
So, myeah...

Did bench marks under linux means SFA to most people.
People into PC wants raw performance on windows.

The fact of the mater is bulldozer could be a 100 core CPU
and still fail to deliver performance as it sucks compared to sandy bridge.
It doesn't matter that is has 8 cores it still sucks. Although it may be cheap other than that if you want performance buy intel. Or AMD server chips maybe? Can you even buy them in the shops?


Maybe AMD should Aim at being 80% as efficient as ivy bridge.

First think they should do is fire all there marketing staff.
Rename the FX to the Crap X or Cyrix .... Or just phenom 1.2

Then pull there head out of there backside and make something that is better than the phoenm2 its sad the FX isn't half as good as the phenom 2

 

[librin.so.1]

pantherx12, eidairaman1 Thanks for clearing that up!

Yeah, I know well enough about those things.
Just from past experience, where companies used flashy (and highly indicating) names for simple things, just to attract as much attention from simple users as possible...

...Simply put, I thought this is one of those flashy sounding things that basically fiddle with the clocks in some clever way, which would prolly be frustrating to OCers.

So, Thanks for saving me the trouble of long and boring Googe'ling (actually DDG'ing) session, and simply clearing things up for me! ;]

 

[librin.so.1]

Did bench marks under linux means SFA to most people.
People into PC wants raw performance on windows.

Oh, please...
To me is the reverse - most of my processor taxing tasks, and furthermore, processing intensive work & study related things are on Linux.
So personally, bought bulldozer without too much thought because I hardly cared if it is slow on windows according to reviews, as it is fast on linux, according to reviews.

As much as I encountered, people using linux in a serious way usually stay in their own communities and do not venture into places such TPU very often. Yet, they are quite numerous. Thus, I suppose You should fix that into "Most simple users into PC want raw performance on windows." ;]

 

[techtard]

AMD fans should learn from the Bulldozer/FX launch. Don't get your hopes up. Too much hype will start another massive troll fest and flame war.

For most average computer users AMD CPUs are perfectly fine, all they need to do is power the OS, web-browser and e-mail client.

'Enthusuiasts' need to realize that they are not the only computer users, and not everyone needs the absolute best PC.

 

[Baam]

I will be upgrading to this for sure.

 

[TheoneandonlyMrK]

Come on take my fin money already.....

these should perform closer to sandy and ivy due to Fm3 support also i hope

Bulldozer is not significantly powerful than Phenom II, and the average user cannot take advantage of the superior multitasking by Bulldozer (they do fine in server though).

beguiles me when i here this?

im the average user ,my rigs listed ,my pc is folding for Tpu in the chimpchallenge(last ditch promotion) and is running 1520 threads at the minute, not gameing is doing my nut by the way, ive not been off ere at all>

Then pull there head out of there backside and make something that is better than the phoenm2 its sad the FX isn't half as good as the phenom 2

its sad someone taught you to use a keyboard