The official AMD Piledriver Discussion Thread

[gmaster456]

I think a lot of people here will only measure it against Intel's best, if it is still slower than a 4.8GHz Sandy/Ivy, then it'll be a failure. We can just about guarantee it will be slower than an overclocked Intel chip. But, I think that if it can improve on some of Bulldozer's weaknesses in a meaningfull way, then it may yet be a decent chip for the masses. It may even find its way into enthusiast setups if it can clock high enough (wouldn't hold my breath on that, but you never know).

Well said.

 

[gmaster456]

I'm expecting for AMD to more or less match SB in terms of overall performance. It will be worse than SB in some things and better for others. That's at least what I'm hoping for.

 

[lakedude]

Intel has superior fab's => lower node tech (ei. 14nm vs 22nm ect), their always a step ahead =>
better performance/watt.

I would personally love to see FinFET technologies (ei. Intel's 3D trasistor) used on AMD cpus as well... since it gives such a huge reduction in power use.

Yeah, Intel is a step ahead on the process node and Intel has the cool FinFET "3D" technology but neither of those are the real problem for AMD.

Case and point: Intel's Sandy Bridge and AMD's Bulldozer are both @ 32nm and neither have "3D" transistors, yet Sandy Bridge destroys Bulldozer for most desktop applications. It is mostly the design, not the process node.

Sure Ivy Bridge is nice with lower power consumption but most would not squawk too loud if you gave them a 32nm Sandy Bridge 2600k. If you give me a 2600k I'll say "Thank You" and smile, a lot.

 

[georgec84]

How much of an architecture redesign is Piledriver compared to BD?

 

[guskline]

How much of an architecture redesign is Piledriver compared to BD?

From what I have read, Piledriver is not a huge"redign", rather a refined Bulldozer. Resonant Mesh technology will be implemented to improve the power draw. The base clock will be upped to 4Ghz from 3.6 (8350 Piledriver vs 8150 Bulldozer) and other minor tweaks. However, if AMD is lucky the combination may result in a decent 8 core performer that has a little more OC room due to better thermals.

As for "spanking" an I5 and beating a 2600k, I wouldn't bet on it. If the 8350 is near the 2600k in performance and is priced @$225 it will sell like mad.

 

[crazymonkeyzero]

Just curious. How much would you have to overclock the AMD FX 8150 to get performance roughly equal to the 2600k in all aspects? Or are they so far apart no viable overclock (excluding LN2) can close the gap between them?

 

[AtenRa]

Just curious. How much would you have to overclock the AMD FX 8150 to get performance roughly equal to the 2600k in all aspects? Or are they so far apart no viable overclock (excluding LN2) can close the gap between them?

In some multithreaded applications the FX8150 is close to Intel Core i7 2600K.

http://www.anandtech.com/show/4955/the-bulldozer-review-amd-fx8150-tested/7

In some others you have to OC at 4.0 to 4.2GHz to equal the 2600K.

Im expecting an 8-core Piledriver at $200-$220 to be faster and have the same power consumption(completion times) with the Core i5 2500K. At 4GHz it will be faster in some multithreaded apps even than Core i7 2600K.

It will still loose to the Core i7 3770K in power consumption but it will be close in multithreaded performance in most cases.

 

[guskline]

AtenRa: If your prediction for the Piledriver is true (hope it is) it will fly off suppliers shelves.

 

[ShintaiDK]

Just curious. How much would you have to overclock the AMD FX 8150 to get performance roughly equal to the 2600k in all aspects? Or are they so far apart no viable overclock (excluding LN2) can close the gap between them?

Excluding AtenRas handpicked benchmarks. Alot.

Even 50% OC wont get you there.

 

[NTMBK]

I'm largely interested in if they can get the production costs down. If they were able to start hitting hard on price/performance again, then they could make a big dent in the midrange and low end market- you know, the majority of PCs sold. 😉 I don't care if the top end PD is slower than the top end IB, if it's priced the same as an IB i3.

 

[ShintaiDK]

I'm largely interested in if they can get the production costs down. If they were able to start hitting hard on price/performance again, then they could make a big dent in the midrange and low end market- you know, the majority of PCs sold. 😉 I don't care if the top end PD is slower than the top end IB, if it's priced the same as an IB i3.

I doubt. Chips are big, volume is low and 3rd party foundry wants its share too.

 

[AtenRa]

Excluding AtenRas handpicked benchmarks. Alot.

Even 50% OC wont get you there.

The majority of people here are Gaming in 1080p and above 😉

http://atenra.blog.com/2012/06/08/amd-fx8150-vs-intel-2500k-1080p-dx-11-gaming-evaluation/

 

[guskline]

Speaking as one who has 2 2500k rigs for gaming, it would be great to have an 8 core CPU for solid multitasking that was in the $200-225 range that almost equals a 2600k. From what I've read, Bulldozer doesn't fit the bill unless extremely OC'd. At that state power draw is ridiculous.

I'm praying that PileDriver is at least to Bulldozer what Phenom II was to Phenom I. If AMD pulls that off they have a winner.

 

[pelov]

Just curious. How much would you have to overclock the AMD FX 8150 to get performance roughly equal to the 2600k in all aspects? Or are they so far apart no viable overclock (excluding LN2) can close the gap between them?

It depends on the benchmark, but somewhere in the 5.5ghz range to equal a 2600K in single-threaded performance. As far as multi-threaded workloads go, it' would depend on the thread scheduler (assigning 4-and-under to single modules would help) and the higher the thread count in a given application the better off the 8150 would be. Even in single-threaded scenarios it would differ in the given IPC for the application in question, though. I guess the most accurate answer would be: you have to overclock it by a lot and sometimes less 😛

The majority of people here are Gaming in 1080p and above 😉

http://atenra.blog.com/2012/06/08/amd-fx8150-vs-intel-2500k-1080p-dx-11-gaming-evaluation/

You have a very funny GPU benchmark. Why did you use two CPUs?

 

[Sweepr]

Excluding AtenRas handpicked benchmarks. Alot.

Even 50% OC wont get you there.

Its interesting that even at 1080p SB is miles ahead of BD in CPU-intensive games.

 

[Chiropteran]

Excluding AtenRas handpicked benchmarks. Alot.

Even 50% OC wont get you there.

Yeah, not in old single-threaded game engines.

Luckily, those old games are still perfectly playable, 77 fps vs 150? Who even cares?

If more new games can support up to 8 threads like bf3, Piledriver could do well. If they can't, oh well. Crappy console ports aren't going to shine on AMD's CPUs. But again, most of the time these games that won't run well are still well above 60 fps, so... yeah.

 

[ShintaiDK]

Yeah, not in old single-threaded game engines.

Luckily, those old games are still perfectly playable, 77 fps vs 150? Who even cares?

If more new games can support up to 8 threads like bf3, Piledriver could do well. If they can't, oh well. Crappy console ports aren't going to shine on AMD's CPUs. But again, most of the time these games that won't run well are still well above 60 fps, so... yeah.

Assuming 2 things:
1: That you can actually multithread, and beyond 4. If you note BF3, threading first shines at the MP.
2: That you dont depend on a faster main worker thread.

If any of those 2 fails, any design like BD fails badly. And that funny enough basicly happens all the time.

Who cares? basicly you are advocating to buy a CPU for the past. And BD cant even keep 60 FPS in games today. Take a SC2 MP game, and BD is directly horrible

 

[Chiropteran]

Who cares? basicly you are advocating to buy a CPU for the past.

No I'm not, I'm advocating bench-marking for the future. You are the one pulling out benchmarks for 8 year old games to show how horrible bulldozer is. At least use current stuff.

 

[pelov]

If any of those 2 fails, any design like BD fails badly. And that funny enough basicly happens all the time.

No I'm not, I'm advocating bench-marking for the future. You are the one pulling out benchmarks for 8 year old games to show how horrible bulldozer is. At least use current stuff.

Most new PC games are still on a 10-year-old DX9 API, so "new" stuff is 10 years old; this also limits the thread count, btw. Still, Chiropteran is right. Because most games are still DX9, most CPUs will do quite fine and won't get bottlenecked and certainly won't dip you down to below playable FPS. It really doesn't matter if Intel or AMD put out a new chip that's weaker than their last generation, because as far as gaming is concerned it really makes absolutely no difference. It's sort of why VirtuMVP is so stupid, who cares if you get +20% FPS when you're already over 60? Other than bragging about it on forums to stupid people who think that actually matters.

 

[The J]

I thought someone on here had said that there were statements in the AMD Compiler Guide (or whatever it's called) that suggested Vishera would have additional tweaks beyond Trinity, aside from L3 cache. Am I mistaken?

I could see myself getting a Piledriver CPU in the future if it were a decent upgrade from my PhII 955 @ 4GHz, so here's hoping it can at least do that.

 

[thilanliyan]

I really hope power consumption goes down with a performance increase. If it does, I will buy PD since I bought a AM3+ motherboard.

BD consumes way too much power for the performance it gives.

 

[Maximilian]

Most new PC games are still on a 10-year-old DX9 API, so "new" stuff is 10 years old; this also limits the thread count, btw. Still, Chiropteran is right. Because most games are still DX9, most CPUs will do quite fine and won't get bottlenecked and certainly won't dip you down to below playable FPS. It really doesn't matter if Intel or AMD put out a new chip that's weaker than their last generation, because as far as gaming is concerned it really makes absolutely no difference. It's sort of why VirtuMVP is so stupid, who cares if you get +20% FPS when you're already over 60? Other than bragging about it on forums to stupid people who think that actually matters.

60 FPS, a chopper crashes into a building in front of you which then partially collapses, it dips to 30 FPS and gameplay gets slow

90 FPS, a chopper crashes into a building in front of you which then partially collapses, it dips to 60 FPS and gameplay stays the same

 

[pelov]

60 FPS, a chopper crashes into a building in front of you which then partially collapses, it dips to 30 FPS and gameplay gets slow

90 FPS, a chopper crashes into a building in front of you which then partially collapses, it dips to 60 FPS and gameplay stays the same

You don't understand how VirtuMVP works then. The only performance gains achieved were under situations where you don't see spikes, and with it enabled the spikes were more severe. It doesn't raise the ceiling for your FPS, it just makes it look like your rig is performing better than it actually is. Essentially, it does absolutely nothing other than allow a few idiots to gloat about how big their e-penis is.

Secondly, unless you're running a multi-monitor setup or playing a very very CPU intensive game that's poorly programmed (like ARMA 2 or Red Orchestra) or highly CPU demanding (BF3 in multiplayer), you won't see a difference in gameplay between an older CPU and a $1000 3960X. Most games are console ports, thus the hardware really doesn't make as big of a difference as it once used to, particularly on the CPU side as game developers have been leaning far more heavily on the GPU.

In terms of how Virtu functions, it is important to understand the concept of being able to perform what was mentioned on the previous page—being able to manipulate what the GPU does and what it does not do. The underlying technology of Virtu is that the environment is virtualized. This means that instead of the GPUs working on top of the operating system, Virtu adds in a middle layer between the operating system and the GPU. This way, Virtu can manipulate everything that the operating system wants to say to the GPU, and vice versa, without either of them knowing that there is a middleware layer.

The second is most important perhaps, which I will go into later. Virtual V-Sync and HyperFormance will only make a difference in the following circumstances:

a) You suffer from visual tearing in your games, or you actively use V-Sync
b) If your setup (screen resolution and graphics settings) perform better than your refresh rate of your monitor (essentially 60 FPS for most people). If you have less than this, then you will probably not see any benefit.

As for a), nVidia has already has a better solution with their adaptive V-sync on the new Kepler series cards, and
b) basically states that when you want it to perform better it won't because it's not meant to. In short, it really does next to nothing. This isn't surprising given that nobody has spoken a word of it since it was initially tested and that was in the pre-release stages when the motherboards weren't available to the public.

 

[Maximilian]

You don't understand how VirtuMVP works then. The only performance gains achieved were under situations where you don't see spikes, and with it enabled the spikes were more severe. It doesn't raise the ceiling for your FPS, it just makes it look like your rig is performing better than it actually is. Essentially, it does absolutely nothing other than allow a few idiots to gloat about how big their e-penis is.

Yeah my post was aimed at your anything over 60 FPS is pointless comment.

 

[pelov]

Yeah my post was aimed at your anything over 60 FPS is pointless comment.

It'd still apply, although I suppose it's an issue of clarity. When I say "anything over 60 FPS" I also mean the spikes as well. If you're not dipping below the Hz-dependent FPS target, it really doesn't matter how high you go. It's only in the extreme ends of the spectrum like 120hz monitors, 3D gaming, multi-monitor gaming and the occasional highly CPU dependent game where a beefier CPU helps, because nearly all of the time you're better off spending the extra money on a better GPU or monitor or SSD than you are on a CPU that'll give you .5-1 frame more per second. It's the reason the 2500K was Intel's best-selling chip and not the 2600K or the 2011 chips. 2500K isn't just enough, it's more than enough.