Piledriver ES benchmarks

I really wish some of you would clarify your 'IPC' statements a little more. Instructions-per-clock is just that, and it varies greatly based on what you're using to measure it. It's also not relative to the number of cores a CPU has, and remember, BD is a 'quad-core' design with hardware dual-threading for integer code. It's still just a 'quad-core' when it comes to float.
 
Damn seems that Benchmarks are a new religion for some poor sods.
 
bwang said:
If you don't measure performance with numeric benchmarks, what do you measure it with?
Click to expand...

He never mentioned measuring performance. I think he's saying benchmarks aren't everything. For example SSD's. There are multiple different ones, you can't run tons of benchmarks to see which one is the absolute fastest, but you probably couldn't tell a difference between most of them. I got a dual core trinity laptop and it's everything I want in a laptop. I don't need something that destroys everything else in benchmarks if all I'm doing is surfing the web and watching videos.
 
Wow, troll time.

Those benchmarks look like a quick fake.

Benchmark numbers are all over the place. Cinebench and 3DMark report vastly different CPU frequencies.

Cinebench claims the chip runs with an odd multiplier 16,5 x 200,7 is approximately the reported cinebench frequency.

CPU-Z show 8 core 8 thread while Cinebench shows 4c/8t.

CPU-Z runs as 32 bit on a 64 bit Windows.

And there is probably more.

Give it a rest guys and wait until actual review are out and stop bitching over AMD and Intel, pick you poison, stick with it and let other people use whatever they want.

EDIT:

They have both 32 and 64 bit CPU-Z screenshots, guess that moves the fake meter up a notch.
 
Last edited:
mufdvr3669 said:
He never mentioned measuring performance. I think he's saying benchmarks aren't everything. For example SSD's. There are multiple different ones, you can't run tons of benchmarks to see which one is the absolute fastest, but you probably couldn't tell a difference between most of them. I got a dual core trinity laptop and it's everything I want in a laptop. I don't need something that destroys everything else in benchmarks if all I'm doing is surfing the web and watching videos.
Click to expand...

Pretty much. I see far too many people getting so over excited (and not in a good way) over benchmarks that don't actually really mean anything important.

They are just graphs and largely are totally irrelevant in everyday use.

Shame people don't get more challenging over more important issues that face us today. I guess like me they will find out when they get older. What a waste.
 
The people who look at benchmark numbers are not interested in 'every day use'. I myself, play games and create 3D artwork, cinebench is a great example of CPU performance benchmarking that relates to my uses.
 
If those benches are real, i would not be surprised if they will delay the launch of these chips and perform last minute fixes. Maybe the results are awful again and that is why they brought Jim Keller as Chief of Processor Group. I think they will ditch the whole design and start from scratch after vishera piledriver is released.
 
Shark974 said:
Hey, it worked for Hardocp! They said Nvidia cards just "feel" smoother in a recent review :rolleyes:
Click to expand...

This was - I believe - specifically related to SLI vs. CFX, not for single cards, and having used both I agree with them.

We just don't have a good way to measure what is going on (yet)
 
IdiotInCharge said:
I really wish some of you would clarify your 'IPC' statements a little more. Instructions-per-clock is just that, and it varies greatly based on what you're using to measure it. It's also not relative to the number of cores a CPU has, and remember, BD is a 'quad-core' design with hardware dual-threading for integer code. It's still just a 'quad-core' when it comes to float.
Click to expand...

IPC? Iterations per cycle?
 
IdiotInCharge said:
I really wish some of you would clarify your 'IPC' statements a little more. Instructions-per-clock is just that, and it varies greatly based on what you're using to measure it. It's also not relative to the number of cores a CPU has, and remember, BD is a 'quad-core' design with hardware dual-threading for integer code. It's still just a 'quad-core' when it comes to float.
Click to expand...

The "IPC" abbreviation is a little bit misused. The actual "Instructions per Clock" is a theoretical number and not directly relatable to real world (or benchmark) performance.

What we are doing is simply choosing the benchmark of our choice, and dividing by the clock speed to see the relative performance per clock cycle in that particular benchmark.

Really, what we should be calling it is "Performance per Clock (Cycle)", but since we often divide by Ghz and not hz, maybe it should be called "Performance per Billion Clock Cycles", or PPBCC or PPBC?
 
Zarathustra[H];1039022279 said:
The "IPC" abbreviation is a little bit misused. The actual "Instructions per Clock" is a theoretical number and not directly relatable to real world (or benchmark) performance.
Click to expand...

You run single threaded spec, and then compare your IPC vs a large number of other well-known entries.
 
Zarathustra[H];1039022279 said:
The "IPC" abbreviation is a little bit misused. The actual "Instructions per Clock" is a theoretical number and not directly relatable to real world (or benchmark) performance.

What we are doing is simply choosing the benchmark of our choice, and dividing by the clock speed to see the relative performance per clock cycle in that particular benchmark.

Really, what we should be calling it is "Performance per Clock (Cycle)", but since we often divide by Ghz and not hz, maybe it should be called "Performance per Billion Clock Cycles", or PPBCC or PPBC?
Click to expand...

You're spot on; my comment/complaint was for those saying crap like 'OMG BD has 50% the IPC of SnB!!!1. We know that's wrong, but yes, it does depend on the benchmark.

And crap like calling BD an eight-core CPU, etc; which wreaks of fanboy/ignorance. Especially when we know that the top-end BD mostly keeps up with quad SnB, which is it's goal to begin with.

Please guys- we know BD's faults. AMD strictly tuned it for server usage in the first iteration, and that's the very best decision they could have made. That's where the margins are. It only had to be 'fast enough' on the desktop. We know they have cache issues that could dramatically improve single-threaded performance if properly resolved, and we know that GF's 32nm SOI process needs some refinement.

If PD comes on the heals of addressing those problems, we could be looking at a CPU that should eclipse SnB by 10% or more (my hope!) at similar clock-speeds and power draws/heat outputs. And hell, if they can get 32nm SOI under control, it should overclock even better!

Here's to a 6GHz PD quad-module to replace the 2500k/3570k in gaming builds!
 
IdiotInCharge said:
If PD comes on the heals of addressing those problems, we could be looking at a CPU that should eclipse SnB by 10% or more (my hope!) at similar clock-speeds and power draws/heat outputs. And hell, if they can get 32nm SOI under control, it should overclock even better!

Here's to a 6GHz PD quad-module to replace the 2500k/3570k in gaming builds!
Click to expand...

For single threaded performance (well, even for multithreaded performance, up to 4 cores) I feel a 6Ghz overclocked Piledriver core would just barely keep up with a stock 2500K/3570k unfortunately.

If the 2500K/3570k owners overclock (which they probably will, as that's why they bought the more expensive K model in the first place), PD will be pretty far behind again.

My hope for a best case outcome of the PD launch is that - while still being beaten by Intel for this type of application - they will be fast enough to run all applications I run satisfactorily.

In order to do so, they don't have to beat Intel, and I'd still consider an AMD CPU for my next rig (provided it isn't a step backwards for me). I just feel that any expectation of AMD surpassing Intel with PD is rather extreme wishful thinking.

I don't think AMD has a chance of surpassing Intel's x86 performance in Pile Driver, Steam Roller or Excavator for that matter. They are going to need a completely new generation of architecture in order to pull something like that off, and even then, I don't see it happening unless Haswell turns out to be Intel's next P4 fiasco.

In order for AMD to catch up to Intel on raw X86 performance at this point, they need another Intel failure.
 
So in 8 threaded aps it will beat Nehalem cpus ?

Great job AMD now only a generation or two more and we might have something matching i5 750 in single thread performance ;)
 
runs2far said:
Wow, troll time.

Those benchmarks look like a quick fake.

Benchmark numbers are all over the place. Cinebench and 3DMark report vastly different CPU frequencies.

Cinebench claims the chip runs with an odd multiplier 16,5 x 200,7 is approximately the reported cinebench frequency.

CPU-Z show 8 core 8 thread while Cinebench shows 4c/8t.

CPU-Z runs as 32 bit on a 64 bit Windows.

And there is probably more.

Give it a rest guys and wait until actual review are out and stop bitching over AMD and Intel, pick you poison, stick with it and let other people use whatever they want.

EDIT:

They have both 32 and 64 bit CPU-Z screenshots, guess that moves the fake meter up a notch.
Click to expand...

4Core 8 threads is correct.It shows it like this after win7 is hotfixed for bulldozer.
 
If PD comes on the heals of addressing those problems, we could be looking at a CPU that should eclipse SnB by 10% or more (my hope!) at similar clock-speeds and power draws/heat outputs. And hell, if they can get 32nm SOI under control, it should overclock even better!
Click to expand...

There is a chance for AMD to take a clear lead in multithreded benchmarks (like was expected with bulldozer) with PileDriver. However there is almost 0 chance that AMD will be ahead in tasks that 1 to 4 threads. There is also 0 chance AMD will be at the same power / heat outputs for similar performance. Their fabrication process is ~18 months behind combined with a lower IPC / higher clocked + more cores design.
 
Last edited:
drescherjm said:
There is a chance for AMD to take a clear lead in multithreded benchmarks (like was expected with bulldozer) with PileDriver. However there is almost 0 chance that AMD will be ahead in tasks that 1 to 4 threads. There is also 0 chance AMD will be at the same power / heat outputs for similar performance. Their fabrication process is ~18 months behind combined with a lower IPC / higher clocked + more cores design.
Click to expand...

Sounds about right.

I am a ever hopeful that AMD will come out with CPU's that have good x86 per thread performance, but we have to be careful about getting too enthusiastic here, as that will just lead to disappointment when reality sets in.
 
Zarathustra[H];1039023243 said:
Sounds about right.

I am a ever hopeful that AMD will come out with CPU's that have good x86 per thread performance, but we have to be careful about getting too enthusiastic here, as that will just lead to disappointment when reality sets in.
Click to expand...

The real improvements will come with steamroller and its about a year away from now.
Don't know if they will attempt to do it sooner. They are showing that the focus has shifted and i don't know how much they will attempt to do on high end cpus. My personal opinion is they have to keep trying harder and on high end cpus, they have no choice to keep going cause if they stop it will affect the rest of their products. That is my personal view.
 
all i read is blah blah blah, more blah blah blah.

discussion over Chinese benchmarks from a crappy ES aka a low wattage processor running at less than ideal clock speeds.


seriously folks find something better to talk about.
 
ebduncan said:
all i read is blah blah blah, more blah blah blah.

discussion over Chinese benchmarks from a crappy ES aka a low wattage processor running at less than ideal clock speeds.


seriously folks find something better to talk about.
Click to expand...

What do you mean by better? There is "nothing" to talk about AMD cause there is "nothing" there for AMD. You are always supporting amd stuff, even when they fuckup.
Now you don't want to talk about another crappy ES? :D

You should instead criticize them for releasing crappy processors.
 
Saving my official comment for an official report. Also it's to be stated exactly how whatever system they used for the "this test" is configured(if it was even really tested; smells like troll dung over there....). Especially the DRAM Driving, HT, and NB Freq's. Proper config i.e OVER 2200 HT/2600 NB yields significant gains in fps and stability as well as lowering load and unload times. I've been following this "debate" for years. Benches can't tell you real-world performance only a relative point of performance-relevant to the bench itself which can hinder certain hardware while bolstering another. And to overcome the low instruction cycle try upping the FSB on ur Phenom II-based chip to about 280-300Mhz and start configuring your system afterwards.

P.s. I have AMD and INTEL systems. lb for lb the real world differences are nil; not enough to warrant paying an extra 70$ to brag about single-bench performance which, imho is a thing of the past. Add that 70$ to an SSD, raid controller, or better GPU and alleviate some bottlenecking because in essence that's what limits performance. AMD is moving forward while INTEL clutches the limelight and learns from their competitors innovations and mistakes. Do we have to go over the many generations of failed INTEL offerings which used to release PRIOR to AMD cpu's. Now we see INTEL steadily announcing "new chips" after "the other chip's" already been there. Regardless of raw sales some1 has always been playing "catch-up". I'll comment about performance after the chip's in my MOBO :)
 
LOL people arguing over new stepping bulldozer benchmarks.

THESE ARENT PILEDRIVER BENCHMARKS.

Look at CPUZ people. Says Zambezi. Also shows a C0 stepping. All bulldozers shipped now say B2 or something.

Thats why theres barely an increase. Its still a crappy processor
 
It was overpriced, but crappy well that's a bit harsh..over hyped yes.
Let me put it like this Intel i3 or FX6100 for multi threaded work both cost about the same..
How crappy is that?

FX stomps the i3 in that price range and I call that a good deal ;-)

I will wait for proper reviews with the retail processors rather than ES and stuff like that.
 
Mr Spocko said:
It was overpriced, but crappy well that's a bit harsh..over hyped yes.
Let me put it like this Intel i3 or FX6100 for multi threaded work both cost about the same..
How crappy is that?

FX stomps the i3 in that price range and I call that a good deal ;-)

I will wait for proper reviews with the retail processors rather than ES and stuff like that.
Click to expand...

That wasn't the original price point though. That's what people are reasonably willing to pay for it.
 
Judging by our current market I think both companies are plateauing. While INTEL has been gaining enormous performance boosts thus satisfying our "now" need. AMD's approach seems future forward. I give both manus. credit. In the mobile market; which eventually will be the larger of the two(home & mobile) because mobile can be used at home and on-the-go. Conversely, the desktop, well, is slowly phasing into a smaller form factor using less power and performing just as well as a larger "box". Full-tower w/all the bells and whistles sub 100$?? As far as desktops AMD already stated it's all set w/that market; done well in it and is moving forward. They're offering upgrades to current AM3 system owners as a way of "sticking to their word" so-to -speak. They did promise that AM3 mobo users wouldn't be forgotten and they held true to that. This offers s938 owners some longevity while the next-gen(FM sockets) are developed. INTEL is clearly still focused on desktops at the moment but you can bet they're watching their rival closely.
 
PcGuy78 said:
Judging by our current market I think both companies are plateauing. While INTEL has been gaining enormous performance boosts thus satisfying our "now" need. AMD's approach seems future forward. I give both manus. credit. In the mobile market; which eventually will be the larger of the two(home & mobile) because mobile can be used at home and on-the-go. Conversely, the desktop, well, is slowly phasing into a smaller form factor using less power and performing just as well as a larger "box". Full-tower w/all the bells and whistles sub 100$?? As far as desktops AMD already stated it's all set w/that market; done well in it and is moving forward. They're offering upgrades to current AM3 system owners as a way of "sticking to their word" so-to -speak. They did promise that AM3 mobo users wouldn't be forgotten and they held true to that. This offers s938 owners some longevity while the next-gen(FM sockets) are developed. INTEL is clearly still focused on desktops at the moment but you can bet they're watching their rival closely.
Click to expand...

And since when has Bulldozer, with its huge die and horrendous TDP, been a mobile-optimized processor?
AMD's architecture is forward-looking - they know that processors can't be tweaked for improved performance indefinitely, so they're looking to clock speed increases to improve single-threaded performance, and hoping that GPU's will take on the bulk of the FP work.
It kind of feels like the Phenom...AMD releases a quad-core die with unified L3 cache, on-board memory controller, and high-speed point-to-point interconnects, but the manufacturing process resulted in a CPU clocked too low to be competitive. A year later, Intel releases Nehalem - quad core, small L2 caches, big L3 cache, on-die memory controller, and a point-to-point interconnect. But they had the manufacturing capabilities to implement it right.
 
drescherjm said:
Only in highly threaded loads which you probably should not be using on either chip.
Click to expand...

That's a joke right?
So I should not be doing video converstions or using Lightroom for photo work on the FX6100?

Dude get real the FX6100 works just great for this type of work...why is it ill informed people think you have to spend £200 on a CPU to do decent multi threaded work?

I only know this..the AMD offering is a much better choice for heavy threaded workloads. And I would argue I need that peformance in heavy threads than I do with some newbie itunes benchmark. That's where the i3's fall down they just can't handle intensive workloads in heavy threading applications.
 
Condition said:
Is AMD just too far behind the curve to catch up without some kind of miracle happening?
Click to expand...

"some miracle" is the only way AMD has ever been able to compete throughout the history of the company.

Before the Athlon in 1999 they were behind as they are now making anywhere from licences knockoffs of Intels 486 chips and serving as an overflow contract manufacturer for Intel to making their own 586 and k6 chips which were unable to compete.

With the Athlon they hired a brand new team, many of which were part of Digital's CPU team. (They even licensed the DEC Alpha bus). With this they were able to catch up with the Pentium 3, but were set to fall behind again in the next generation.

And then a miracle happened.

Intel totally screwed the pooch with Netburst/P4 allowing AMD to take the lead for quite a few years. They reinvested R&D spending in their CPU designs, and came out with the first consumer 64bit CPU's and their own new instruction set, allowing them to maintain the lead until Intel finally was able recover from the P4/Netburst disaster with the Core 2 CPU's in 2006.

AMD Could have reinvested the profit from this period in cementing their competitive status with Intel by bringing out competing CPU designs. Instead they spent $5.6 billion on acquiring ATi. With this one decision they doomed their CPU future as they simply no longer had the cash to do the type of R&D on the CPU front they needed in order to remain competitive with Intel.

Instead of a strong competitor to Intel's Core 2 architecture we thus got a much delayed and under performing Phenom CPU with a critical hardware bug in 2007, after Intel had been dominating for a year with Core2. They simply didn't have the cash to bring a good CPU to market.

Similar under-investment in CPU R&D led to a lackluster Phenom II and a disappointing Bulldozer.

As it stands today, thus the situation on the CPU front is much like it was in the mid to late 90s before the Athlon was launched. The CPU race is Intel's to lose, not AMD's to win. Without a major screw up on Intel's part AMD will likely not be able to catch back up again.

All this could have been very different if AMD had not decided to buy ATi.

So was it worth it? Was buying ATi and getting into the graphics business worth sacrificing the future of their CPU business? Maybe. It all depends on how well their APU's turn out long term, and how well Intel can improve their on chip HD graphics.

Maybe they thought that they could never win a CPU race long term with Intel, a company 10 times their size and financial might, able to throw armies of well paid engineers at any problem. even if they tried, and thus thought it was wiser to invest in ATi. I have no idea.

I'll leave this for people who know the industry better than I do to determine.
 
"some miracle" is the only way AMD has ever been able to compete throughout the history of the company.
Click to expand...

not exactly. Amd does weill with the resources they have. Intel is just more likely to succeed due to their larger budgets. However both companies are limited by there creativeness .

Amd plans HSA, which honestly fits perfectly for The FX line up of processors. Steam roller is set to the be the first iteration of that.

Amd bought ATI, because it was a solid company, and their graphics expertise would pay off for them in the future, for not only HSA, but discrete graphics cards for Super computers, gamers, and workstation markets.

Would it be a miracle if Amd introduced Steamroller, with HSA, paired with a good graphics card sharing memory. Cpu +GPU working in conjunction with one another, and blasts Intel? I wouldn't call that a miracle i would call it better planning and more creativeness. Unlikely to happen, but hey we will see.
 
Last edited:
Michaelius said:
So in 8 threaded aps it will beat Nehalem cpus ?

Great job AMD now only a generation or two more and we might have something matching i5 750 in single thread performance ;)
Click to expand...

I have an AM3+ board and am hoping to upgrade to Cyrax performance soon.
 
You must log in or register to reply here.
Back
Top