AMD Piledriver rumours ... and expert conjecture

[Reynod]

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 ...

 

[palladin9479]

For those doubting the usefulness of multiple cores, additional cores will ALWAYS be useful, just not in a way that's easily bench-marked by the enthusiast.

http://en.wikipedia.org/wiki/Context_switching

I have 929 threads running on my work internet accessibly computer. I only have outlook and Mozilla Firefox open with nine tabs. There are a ton of background management / security related tasks running, this is an enterprise environment after-all and the security section is like that. All those threads need time on the CPU eventually, some more often then others. When they need time they interrupt whatever the CPU is doing which forces the CPU to switch states and do a reset / flush. More cores means more simultaneous states can be active.

This is why Sun designed in eight threads per CPU core (HT only does two), to enhance performance on a system with hundreds to thousands of process's working simultaneously. Do not discount the usefulness of having additional cores because a system with everything turned off running a single player timed loop demo showed a 20% better performance on one CPU vs another.

 

[jimmysmitty]

IPC is the key...

Thats a no-no around here. I mean so what if it takes a higher clock speed to do the same work, I mean AMD used to pride themselves on doing the same work in a lower clock speed and thus a lower thermal envelope but that is not what matters now.

I honestly think the $ part only matters per person. Some may need to be in the $100 range, others the $300 range. Back during Phenom II, some people could only budget low price for a CPU and in those cases, AMD looked better for the price.

But there is still the overall performance which still matters in some ways and thats where IPC come in, or the made up term that Intel and AMD both use.

What if i have a 3.0Ghz Cpu and you have a 1.0Ghz CPU with twice the IPC.....No i'm joking with you! I HATE THIS SAYING!! AMDZONE be Damn!

Who ever says this needs to understand that Their is no 1.0Ghz flagship CPU's anymore! Amd only has a 600mhz lead or so. But their cores are only 50% as good as Intel's per clock!

To step up with Intel(Which we all want but know it isn't happening) They would Twice the performance per clock or at least 85% as powerful with a 20% or so higher clock rate which just isn't possible really given Intel's high 3.5Ghz clock already and Amd's approach with the module design which is best for multithreaded apps.


I'm sure their is Engineers here who can tell you guys this just isn't reasonable to have a CPU clock much higher then 4.2Ghz at stock(If that with 6+ cores).

So All Amd can compete with is Graphics and Moar cores. Which I'm quite proud of Intel and its ivy bridge when it comes to the graphics no longer will it be a bottleneck for 1080P video or 3 monitor video for business tasks. And as the future goes on Intel will give users more cores for less money and Amd will start to lose its Moar cores lazy strategy. But i just hope Intel wont catch up to Intel in the graphics department(when it comes to apu's) If so then darn!


Come On amd get to work! Keep taking Market share from Intel!

http://www.itworld.com/hardware/271752/amd-gains-x86-processor-market-share-intel-q1

Intel catching up to Intel? That sounds like a closed loop.... 😛

I hope Intel does catch up to AMD. I hope they push back and forth. Its never good to have only one option, considering that nVidia isn't in the APU business.

And I think Intel will at some point catch up to AMD. A lot of people said Intel could never double the older crappy GMA performance, the first gen Core i GT did. Then they said SB wouldn't be better. It was also about double. Intel has the ability to do it. I think its good that they are pushing this hard. Probably makes AMD sweat knowing they may have a fight on their hands at some point. They probably think back to Core 2 and how it made them feel.

Or at least I hope they have that mindset to keep pushing and not standing by with the same product with very little changes.

 

[jdwii]

Thats a no-no around here. I mean so what if it takes a higher clock speed to do the same work, I mean AMD used to pride themselves on doing the same work in a lower clock speed and thus a lower thermal envelope but that is not what matters now.

I honestly think the $ part only matters per person. Some may need to be in the $100 range, others the $300 range. Back during Phenom II, some people could only budget low price for a CPU and in those cases, AMD looked better for the price.

But there is still the overall performance which still matters in some ways and thats where IPC come in, or the made up term that Intel and AMD both use.



Intel catching up to Intel? That sounds like a closed loop.... 😛

I hope Intel does catch up to AMD. I hope they push back and forth. Its never good to have only one option, considering that nVidia isn't in the APU business.

And I think Intel will at some point catch up to AMD. A lot of people said Intel could never double the older crappy GMA performance, the first gen Core i GT did. Then they said SB wouldn't be better. It was also about double. Intel has the ability to do it. I think its good that they are pushing this hard. Probably makes AMD sweat knowing they may have a fight on their hands at some point. They probably think back to Core 2 and how it made them feel.

Or at least I hope they have that mindset to keep pushing and not standing by with the same product with very little changes.

Opps that's what happens when i type and do Math at the same time!

 

[esrever]

IPC doesn't matter as long as the total performance is there.

Sandybridge can be better than ivybridge even though ivy has better ipc. The simple fact that sandy can clock higher makes up for it.

There are much more factor at play than IPC. Its nice to have decent IPC but only looking at IPC is just failing to look at the larger picture.

If AMD made bulldozer on a mature process I can imagine it clocking much higher, maybe still not having as good single threaded performance as sandybridge but it would probably look much better in comparison than now.

I wouldn't be surprised if piledrive delivers base clocks of 4ghz+ and turbo of 5ghz+.

 

[jdwii]

For those doubting the usefulness of multiple cores, additional cores will ALWAYS be useful, just not in a way that's easily bench-marked by the enthusiast.

http://en.wikipedia.org/wiki/Context_switching

I have 929 threads running on my work internet accessibly computer. I only have outlook and Mozilla Firefox open with nine tabs. There are a ton of background management / security related tasks running, this is an enterprise environment after-all and the security section is like that. All those threads need time on the CPU eventually, some more often then others. When they need time they interrupt whatever the CPU is doing which forces the CPU to switch states and do a reset / flush. More cores means more simultaneous states can be active.

This is why Sun designed in eight threads per CPU core (HT only does two), to enhance performance on a system with hundreds to thousands of process's working simultaneously. Do not discount the usefulness of having additional cores because a system with everything turned off running a single player timed loop demo showed a 20% better performance on one CPU vs another.

Rubbish, Most of those threads are completely idle, Also 90% of programs don't even use 4 cores so why should we have 8 to get the same performance as 4 cores? Sorry but it takes 8 Amd cores at full power to take on 4 Intel cores with less clock speed and a weaker Turbo! IPC MATTERS MORE. How long does Intel need to beat Amd for some users to understand this?????

 

[jimmysmitty]

IPC doesn't matter as long as the total performance is there.

Sandybridge can be better than ivybridge even though ivy has better ipc. The simple fact that sandy can clock higher makes up for it.

There are much more factor at play than IPC. Its nice to have decent IPC but only looking at IPC is just failing to look at the larger picture.

If AMD made bulldozer on a mature process I can imagine it clocking much higher, maybe still not having as good single threaded performance as sandybridge but it would probably look much better in comparison than now.

I wouldn't be surprised if piledrive delivers base clocks of 4ghz+ and turbo of 5ghz+.

We are a minority. We enjoy the finer aspects of PC, including high quality parts, overclocking and maximum performance. But the majority does not. The majority uses stock clocks and normally lower quality parts so IPC is much bigger there as a 3770K is the same clock speed as a 2700K.

BD on the other hand has higher clocks, higher power consumption but lower IPC so it gets the meh opinion overall except for overall lower price per core.

As for the mature process, possibly. But then again it depends on what their clock ceiling is. There will still be a GHz barrier we wont be able to pass for some time, I believe anyways. I don't think we will see the majority of CPUs doing 5GHz on air until they find a new material such as graphine or carbon nano tubes or whatever Intel/AMD/IBM cook up.

PD, I expect higher clocks but not higher IPC. The clocks will do the major performance gains, IPC will not. But thats my opinion. I think AMD wont see a major performance jump until their next arch.

 

[palladin9479]

Rubbish, Most of those threads are completely idle, Also 90% of programs don't even use 4 cores so why should we have 8 to get the same performance as 4 cores? Sorry but it takes 8 Amd cores at full power to take on 4 Intel cores with less clock speed and a weaker Turbo! IPC MATTERS MORE. How long does Intel need to beat Amd for some users to understand this?????

*Cough*

939 * .10 = 93.9 threads that are not "idle". An 8 core CPU would have 11.7 active threads per core according to your statement.

BTW its more like 99% as most theads are waiting for an event to be triggered. You don't get to control when that event is triggered on an active system and there are hundreds of OS level events being triggered every second, you simply don't see them. How do you think that mouse is being moved across that screen? The mouse input is going to a port that's being fed into a memory address as positions, a task then takes those positions calculates any required events "mouse over / ect..." then pass's those coordinates to the GDM / WDM which renders it to the frame buffer. And that's for a single mouse movement done tens if not hundreds of times per second. Every last one of those is a context switch as your CPU isn't only rendering a mouse moving across the screen.

Clock speed means nothing, its performance at a certain cost vs power usage (another word for cost). Ultimately it's performance vs cost*2 vs density.

Simple though experiment,

CPU A at 1 Thz (physics being ignored for the time being) or 1,000 Ghz
CPU B at 1Ghz

CPU A can process 1 woglet (invited metric of performance) per clock cyle and thus has a performance of 1,000,000,000,000 woglets per second or 1 Twog per second.

CPU B can process 1,000 woglets per cycle and thus has a performance of 1,000,000,000,000 woglets per second or 1 Twog per second.

CPU's A and B cost the same and produce the same thermal output.

CPU B has 100,000% more "IPC" then CPU A.

Both CPU's are functionally identical.

I used big numbers to denote that even large differences, ridiculously earth shattering large, in "IPC" mean absolutely nothing in the grand scheme of things. What matters is the performance you get for the cost of the product. Or rather, the method used to pluck the chicken doesn't matter provided the chicken gets plucked.

Welcome to my world, where we build and compare various engineering solutions that best match our customers needs.

 

[truegenius]

Opps that's what happens when i type and do Math at the same time!

so you need multicore cpu (brain) to stop reseting of a thread 😛

M.A.L.MENTAL wrote
'ipc is the key'
i quote
:lol:

so its a need to replace ipc with ppg (performanbe per gigahertz) based on a standard like c11.5

the thing that matters is
ghz x ppg in less tdp and less price
and for multithreading, multiply that by number of cores (assuming full use of cores)

HOW ABOUT USING CPU TIME TO CALCULATE CPU USAGE BY A TASK/PROCESS? 😀

 

[dogman_1234]

I-P-C-K-E-Y

Ipckey, Ipckey House!

 

[Cazalan]

If your cache latency is crap good luck with lots of context switches.

 

[palladin9479]

If your cache latency is crap good luck with lots of context switches.

Pretty much. Although L1 is by far the most important.

I never made any arguments ~for~ AMD, only that more cores are always useful and core count should never be ignored.

 

[jdwii]

Pretty much. Although L1 is by far the most important.

I never made any arguments ~for~ AMD, only that more cores are always useful and core count should never be ignored.

What you said makes since but we can't forget about IPC Since you cant run the CPU's at such a high clock rate anyways, And Most programs will still work better having more powerful cores vs more of them, I'm not talking about servers where TDP per watt(Which Intel has amd beat also) matters the most i'm talking about Overall-Performance/per dollar that's all i care about within reason.

And to let you guys know 1 core can do multitasking as well, Just because their is twice as much cores doesn't mean its better then a processor with twice the performance per core. If anything that means the processor with twice the cores is worse because now it needs to have all its cores in use to take on the processor with half as much cores which isn't always possible.



I really don't care much to have the performance per core of Intel i just want it to be close and if Amd can give me twice the cores as well then its a win win but Currently this is not the case..not even close. I sure hope it will be though unless my sabertooth will cry.

 

[palladin9479]

What you said makes since but we can't forget about IPC Since you cant run the CPU's at such a high clock rate anyways, And Most programs will still work better having more powerful cores vs more of them, I'm not talking about servers where TDP per watt(Which Intel has amd beat also) matters the most i'm talking about Overall-Performance/per dollar that's all i care about within reason.

And to let you guys know 1 core can do multitasking as well, Just because their is twice as much cores doesn't mean its better then a processor with twice the performance per core. If anything that means the processor with twice the cores is worse because now it needs to have all its cores in use to take on the processor with half as much cores which isn't always possible.

Not always, and the reason I keep saying not to use single threaded timed loop demo benchmarks to guide purchasing decisions. Especially if those benchmarks were done with all services, AV and management services disabled.

A single "Core", or rather a single processing target (register stack) is incapable of doing more then one thing at a time. Two instructions can not be executed on a single register stack as all x86 instructions assume they have 100% control over the register stack. In order to simulate doing multiple things at once the CPU task switch's constantly between ring 0 threads. It's those ring 0 threads that control what ring 3 threads (user mode) get executed on the CPU, how much time their allowed to have and what is allowed to interrupt them. Having a second register stack (Hyper Thread) immediately cuts your task switching in half even if it offers no more additional processing power. Adding a second "core" will not only cut your task switching in half but also give you a second set of processor resources that are separate (supposedly) and thus not interdependent with the first core.

Ultimately, "IPC" is a useless metric when comparing different uArchs. I could easily bring up T3 vs T4 chips. T3's run at 1.67Ghz but had 16 cores with 8 threads per core. The T4s run at 2.85Ghz and have 8 cores with 8 threads per core. In raw "IPC" the T3 wins, but in total performance the T4 bypass's it. The reduction in cores along with changes in uArch allowed Sun (Oracle) to run the clock speed up which resulted in better overall performance.

From everything I can see, each BD "core" is 0.75 of what a K10 core is which is actually amazing considering each BD core only has 2 ALU's where each K10 core has 3 ALU's. Thus your getting 75% of the performance with only 66.6% of the processing resources. The performance issues with BD come from the high latency in the L2/L3 cache and the arbitration issues caused by having a 1,1,2 / 1,1,1+1 scheduler vs a 1,1,1,1 scheduler.

Or are you suggesting a 100mhz CPU that runs exactly 1 instruction per cycle is better then a 1Ghz CPU running 0.5 instructions per cycle. After all the first CPU has higher "IPC".

 

[truegenius]

ipc per core

""the arbitration issues caused by having a 1,1,2 / 1,1,1+1 scheduler vs a 1,1,1,1 scheduler.""
check bounced 😀
means i didnt got that

 

[palladin9479]

ipc per core

""the arbitration issues caused by having a 1,1,2 / 1,1,1+1 scheduler vs a 1,1,1,1 scheduler.""
check bounced 😀
means i didnt got that

Every BD module shares two things, first being the L2 cache and the second being the front end instruction scheduler / decoder.

Now there are four instruction scheduler / decoder units on the front end of the module, so two cores sharing four instruction schedulers. What Agner found out was that the front end schedulers are not equal. Two of them are regular schedulers but the third and fourth scheduler is bonded together. So you can only get four instructions per cycle if you have a mix of ALU / FPU or all sinple ALU instructions. If you have two double's (two instructions represented as one) then only one of those doubles can be decoded at a time. If you have four ALU instructions then all four can be processed at once.

Hope this kinda sorta makes sense. It's the part where they take x86 macro-ops and create mico-ops for the internal resources to process. On most code you'll get about 3 decided instructions per cycle for both cores so only 1.5 on average. If two threads both are heavy with FPU instructions then you'll end up with serious arbitration issues and stalled CPU cores. If the two threads are heavily populated with ALU instructions then you'll be fine for arbitration. This validates the findings that you can get better FPU performance by disabling one core per module. Heck you can get better over-all performance for most tasks by disabling every other core. Not the most efficient and your kinda wasting CPU power but if it gets you better performance then go for it. If your board doesn't support core locking / unlocking you can use the el'cheapo method and use NT processor-affinity. Set your process / task to only run on every other core and you'll have much the same result.

On a side note, I've been able to use my OC A8-3530mx "turbo-boost" to great effect using this technique. I'll force a process of a program I'm running to only run on core 0 which immediately spiked to 2.4Ghz and lowers the other three to 800mhz. This is vs Windows NT moving the process around to each of the four cores and keeping them all at 1.9Ghz.

 

[jdwii]

Not always, and the reason I keep saying not to use single threaded timed loop demo benchmarks to guide purchasing decisions. Especially if those benchmarks were done with all services, AV and management services disabled.

A single "Core", or rather a single processing target (register stack) is incapable of doing more then one thing at a time. Two instructions can not be executed on a single register stack as all x86 instructions assume they have 100% control over the register stack. In order to simulate doing multiple things at once the CPU task switch's constantly between ring 0 threads. It's those ring 0 threads that control what ring 3 threads (user mode) get executed on the CPU, how much time their allowed to have and what is allowed to interrupt them. Having a second register stack (Hyper Thread) immediately cuts your task switching in half even if it offers no more additional processing power. Adding a second "core" will not only cut your task switching in half but also give you a second set of processor resources that are separate (supposedly) and thus not interdependent with the first core.

Ultimately, "IPC" is a useless metric when comparing different uArchs. I could easily bring up T3 vs T4 chips. T3's run at 1.67Ghz but had 16 cores with 8 threads per core. The T4s run at 2.85Ghz and have 8 cores with 8 threads per core. In raw "IPC" the T3 wins, but in total performance the T4 bypass's it. The reduction in cores along with changes in uArch allowed Sun (Oracle) to run the clock speed up which resulted in better overall performance.

From everything I can see, each BD "core" is 0.75 of what a K10 core is which is actually amazing considering each BD core only has 2 ALU's where each K10 core has 3 ALU's. Thus your getting 75% of the performance with only 66.6% of the processing resources. The performance issues with BD come from the high latency in the L2/L3 cache and the arbitration issues caused by having a 1,1,2 / 1,1,1+1 scheduler vs a 1,1,1,1 scheduler.

Or are you suggesting a 100mhz CPU that runs exactly 1 instruction per cycle is better then a 1Ghz CPU running 0.5 instructions per cycle. After all the first CPU has higher "IPC".

No i'm not but in today's world their is not a 10X difference in clock speed their is only at best a 30% difference such as the 4170fx vs 2500K. So bad example as well as the T3 vs the T4 as the T4 had a 70% clock speed advantage.

Again people need to understand that this clock speed advantage just isn't reasonable in today's world. The highest i can see Piledriver being clocked at is 4.2Ghz with a 4.7Ghz max turbo and that would be pretty darn amazing almost to the point of WoW! And even that would only be 16.6% advantage over Bulldozer. That would not even be anywhere close enough to close any gaps what so ever.

Its going to take a long time for Amd to fix their problems with their CPU architecture and its not going to work by people(including Amd) blowing off per core performance or IPC when the majority of the apps only use a few cores(remember Intel is giving us 4 cores for the same price as Amd's 8) but tax them very well, such as games. Their is some games that are Skippy on a Amd system even with a 7970, Their is some sites that show the 8150fx as being even with the 2600K even when all 8 cores are being stressed sorry but that's sad given the amount of TDP(125watts) and the amount of extra die space it takes to do so. 1.2B transistors vs 995M for the I7.

Thanks palladin9479 for having a nice conversation, Mods don't get mad about this we are doing no harm to anyone or name calling.

 

[truegenius]

""Windows NT moving the process around to each of the four cores and keeping them all at 1.9Ghz""
means flushing and reseting and loss of performance due to jugling of threads

so amd can use die harvesting to increase gaming performance and to decrease tdp and to get higher clocks

i too think that setting affinity or disabling cores can increase gaming performance

 

[Chad Boga]

No i'm not but in today's world their is not a 10X difference in clock speed their is only at best a 30% difference such as the 4170fx vs 2500K. So bad example as well as the T3 vs the T4 as the T4 had a 70% clock speed advantage.

These crazy exaggerated clockspeed discrepancies being used in theoretical discussions involving IPC, is a continuation of the woeful behaviour which saw gullible and/or young posters on forums all around the world getting sucked into buying an AM3+ motherboard before Bulldozer came out.

We all saw how well that worked, but hey, some people need to be sacrificed, to meet other people's ideological obsessions.

 

[gamerk316]

For those doubting the usefulness of multiple cores, additional cores will ALWAYS be useful, just not in a way that's easily bench-marked by the enthusiast.

http://en.wikipedia.org/wiki/Context_switching

I have 929 threads running on my work internet accessibly computer. I only have outlook and Mozilla Firefox open with nine tabs. There are a ton of background management / security related tasks running, this is an enterprise environment after-all and the security section is like that. All those threads need time on the CPU eventually, some more often then others. When they need time they interrupt whatever the CPU is doing which forces the CPU to switch states and do a reset / flush. More cores means more simultaneous states can be active.

This is why Sun designed in eight threads per CPU core (HT only does two), to enhance performance on a system with hundreds to thousands of process's working simultaneously. Do not discount the usefulness of having additional cores because a system with everything turned off running a single player timed loop demo showed a 20% better performance on one CPU vs another.

No one is arguing more cores are useful. But its important to remember, within the confines of a SINGLE PROCESS, it is not always easy to make the code parallel to the point where you see a benefit of more cores. Games tend to be one such example.

More cores works great when dealing with hundreds of processes, even if non of them can scale. I've always held BD is a great design for a server CPU, where this condition holds true.

More cores works great for multiple processes, but thats no guarentee that a SINGLE process can scale to any degree.

 

[gamerk316]

IPC doesn't matter as long as the total performance is there.

Sandybridge can be better than ivybridge even though ivy has better ipc. The simple fact that sandy can clock higher makes up for it.

There are much more factor at play than IPC. Its nice to have decent IPC but only looking at IPC is just failing to look at the larger picture.

If AMD made bulldozer on a mature process I can imagine it clocking much higher, maybe still not having as good single threaded performance as sandybridge but it would probably look much better in comparison than now.

I wouldn't be surprised if piledrive delivers base clocks of 4ghz+ and turbo of 5ghz+.

For a single core:

Performance = Instruction Per Cycle * Number of Cycles

Speed (Hz) is half the equation, IPC is the other. They count equally.

If CPU x can do twice the number of work per clock as CPU y, then CPU y needs to be clocked twice as high to offer the same performance. Simple. If the CPU is 20% slower in IPC, it needs to be clocked 20% higher to compensate.

[I now stress IPC varies by workload, and other parts of the system come into play. I'm dumbing things down for the purpose of discussion].

Heres the issue: Due to heat/power draw, you don't see large frequency differences in CPUs any more. As a result, IPC is what separates a "good" CPU from a "bad" CPU.

More cores simply act as a multiplyer to the above equation. When using SMT [be it HTT or CMT or some other implementation], you have some sort of scaling factor [10% for HTT, 80% for CMT, etc].

[I now stress this assumes a 100% full load where all the processing power of the CPU is used, which typically isn't true. If it was, BD would look much better then it currently does].

So to compare an i5 and 4000 series BD, you get something that looks like this:

SB i5:
Performance = (IPC * Clockspeed) * 4

40xx BD:
Performance = ( ((IPC * Clockspeed) * 2) + (((IPC * Clockspeed) * 2) * 80%) ) [2 full cores, 2 CMT cores]

Lets assume both are clocked at 3GHz at stock and have the same exact IPC [which we know is false].

SB i5:
Performance = (1 * 3) * 4 = 12

40xx BD:
Performance = ( (1 * 3) * 2) + (((1 * 3) *2) * .8) )
Performance = ( (6) + (4.8) )
Performance = 10.8

Reducing to 1:
BD = 1
i5 = 1.111....

Thats your performance difference. Even if BD and SB (i5) has the same exact IPC, due to the 20% performance hit of using CMT, you'd still be 10% slower overall. And since we know BD has a lower IPC then SB, that difference tends to be larger.

My point being, you can mathematically solve performance differences in any given benchmark to figure out the difference in IPC for that specific benchmark, since the performance difference for the benchmark and speed of the CPU are known. I'd like to see some review sites start to look into IPC differences on a per-benchmark basis, as it would be interesting...

 

[-Fran-]

These crazy exaggerated clockspeed discrepancies being used in theoretical discussions involving IPC, is a continuation of the woeful behaviour which saw gullible and/or young posters on forums all around the world getting sucked into buying an AM3+ motherboard before Bulldozer came out.

We all saw how well that worked, but hey, some people need to be sacrificed, to meet other people's ideological obsessions.

I wonder how many people waited and bought the Z77 MoBos... Hell, even waited for IB.

What you said there, works both ways. But I'm sure you'll twist it and keep the "all the people buying AMD is stupid" mantra. Your sarcasm is getting very annoying.

---

Anyway, instead of talking and screaming about what a single core can do for your lives (IPC), why not think about what palladin said for a moment. IPC is just another metric, but by itself means squat.

Cheers!

 

[Chad Boga]

I wonder how many people waited and bought the Z77 MoBos... Hell, even waited for IB.

What you said there, works both ways. But I'm sure you'll twist it and keep the "all the people buying AMD is stupid" mantra. Your sarcasm is getting very annoying.

---

For it to work both ways, you would need to have had people making outlandish claims for what IB would bring.

You would have needed to have an Intel employee spreading woeful disinformation.

As there was a distinct absence of the above, what you apparently think works both ways, all of a sudden becomes a very definite one way journey.

Anyway, instead of talking and screaming about what a single core can do for your lives (IPC), why not think about what palladin said for a moment. IPC is just another metric, but by itself means squat.

Everyone knows it means nothing in isolation.

Inherent in the discussion is that when the situation involves quite similar clock speeds, sufficient number of cores for the desktop, reasonable pricing, thermals and energy use, then what becomes the paramount distinguishing feature is IPC.

It takes real effort to pretend none of the above was inherent in the argument put forward by people talking about the importance of IPC.

 

[-Fran-]

For it to work both ways, you would need to have had people making outlandish claims for what IB would bring.

You would have needed to have an Intel employee spreading woeful disinformation.

As there was a distinct absence of the above, what you apparently think works both ways, all of a sudden becomes a very definite one way journey.

With one little difference, AM3+ offered a tad more than just "BD compatibility". I'm only focusing, to be fair, on your "AM3+ was a stupid purchase" argument which I find a little biased. But overall, I can see where you're going and seems a fair assessment on the BD side of things. I also fell in that part of the argument and AMD indeed had fault there, but my reason behind the purchase was SLI support.

Everyone knows it means nothing in isolation.

Inherent in the discussion is that when the situation involves quite similar clock speeds, sufficient number of cores for the desktop, reasonable pricing, thermals and energy use, then what becomes the paramount distinguishing feature is IPC.

It takes real effort to pretend none of the above was inherent in the argument put forward by people talking about the importance of IPC.

If it means nothing, then why blame everything in BD's "poor IPC figures"? Most of the arguments behind the "BD is crap" is the poor IPC, when in reality it's just ONE of many factors involved and not the most important as palladin is trying to make you guys notice.

Anyway, thanks for the "correct" way of responding, Chad. Try to lessen the sarcasm a little though, please 😛

Cheers!

 

[Chad Boga]

With one little difference, AM3+ offered a tad more than just "BD compatibility". I'm only focusing, to be fair, on your "AM3+ was a stupid purchase" argument which I find a little biased.

That is not quite my argument.

I was referring to various attempts by various parties to downplay the importance of IPC both before the release of Bulldozer and even more amazingly, after the release of Bulldozer.

Wrapped up in all this, was as of a consequence of dishonest arguing, the unfortunate situation of newbies and/or the gullible, jumping the gun and ending up with severe buyer's remorse.

If you bought for SLI support, then obviously you weren't one of the people negatively affected by this, but many were, including jdwii, whose arguments now about the importance of IPC, are I believe, coming under attack from bad faith arguments that wander off into server land, when the discussion is so clearly desktop focussed.

If it means nothing, then why blame everything in BD's "poor IPC figures"?

It means nothing in isolation. Read what I said is inherent in the argument.

Most of the arguments behind the "BD is crap" is the poor IPC, when in reality it's just ONE of many factors involved and not the most important as palladin is trying to make you guys notice.

Palladin is wrong and it looks like you haven't even understood his disingenuous arguments.

Bulldozer has a clock speed advantage over SB, it has a core advantage, it has pricing, thermals and energy use that one could live with, so where does it fall down against SB, other than clearly on IPC.

 

[truegenius]

but we are still limited by software not able to use more core effectively