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 ...

 

[Dangi]

I still don't see the point of comparing A10-5800K vs i5 and i7.

APUs are targeted to the lower end market so Pentium and i3 is what is targeted in performance and in price, so put A10 agains i3 and pentium that cost almost the same, not against a processor that cost almost triple

A10-5800K 129$
http://www.newegg.com/Product/Product.aspx?Item=N82E16819113280

i7-3770K 330$
http://www.newegg.com/Product/Product.aspx?Item=N82E16819116501


As said in the previous page put and i7 vs dual Xeon and see if i7 is worth buying....

 

[fazers_on_stun]

My gripe about benches is that they rarely ever manifest in real life, the performance difference between intels top chip and bulldozer, faildozer, bullcrap whatever people called it is still not distinguishable, if it were 55% faster it would be noticeable even at the basic level of booting up, responsiveness.

Suppose you went shopping for a new car, maybe a sports car, but there were no reviews comparing different models to each other - all you could get were the basic facts like length, curb weight, engine horsepower and fuel economy. Short of going to 10 different dealers and actually test-driving each model, how would you get the most performance for your $$?

While benchmarks are nowhere near perfect, and YMMV, they remain an effective tool for most people to gauge relative performance so they can make an informed purchase decision according to their wants and needs..

 

[fazers_on_stun]

Why the hell is the A10 benched against an i5 ?!!! How lame and stupid is that? The A10 falls in the i3 price range....no point in comparing it with an i5, then marking everything in red to show that its slower !!

My guess is that it's 4 AMD cores (AMD's marketing once again) vs. 4 Intel cores, although they should have used a 3570K (no HT) in order to get closer stock frequencies..

 

[sarinaide]

^
they were like this
pd/trinity will be equal/near to phenom/k10 in performance per ghz
and more clockable thus better per core performance
but due to performace loss in a module , 4ghz module will be equal/near to 3.2ghz 2xk10 cores in multithreaded

and thus 8350 looks like 20-25 % better than 1090t and 8150 in multithreaded
and ~10% better than 1090t in single thread
everything at stock, with marginal error in prediction 😛 😀

I would have gone along with that 😀 It would have been a good wager.

 

[guskline]

With the release of Trinity we now can see, in a rough way, the improvement of the PileDriver over the Bulldozer. I the tech report I read, which compared the A10 to a 4170 Bulldozer the improvement was as high as 7%. There were clearly limiting factors to the A10 versus the 4170 (ie no L3 cache for the A10 plus F2 socket) However, the PileDriver clearly shows improvement. Just not "Earth shattering". I think sarinaide's thought that SteamRoller will be the big jump in performance is spot on.

 

[Cazalan]

Newegg has the A10-5800K for $124.99 right now. $104.99 for the A8-5600K.

At that price they could win back some sales.
I agree comparing it with anything beyond an i3 is a disservice.

SA comparing it with a $300+ i7-3770K was ridiculous.

 

[truegenius]

does it mean that an i3 is not having enough muscle to be compareable to a10 (in multi/gaming) and thus they brought out the monster i5 but sleepy (locked) 😗

 

[gamerk316]

My guess is that it's 4 AMD cores (AMD's marketing once again) vs. 4 Intel cores, although they should have used a 3570K (no HT) in order to get closer stock frequencies..

I agree on that front, but its not like HTT is going to have a huge effect one way or another. Also worth noting is the clockspeed edge the A10 has...

My point being: In CPU benchmarks, Piledriver is still going to trail Ivy by a bit, though it looks like my 8% prediction was just about right.

 

[truegenius]

http://www.flipkart.com/browse/computers/components/processor-20246?_r=Z3JnWgVY70wSGtZ5oHr1ag--&_l=h2lepvuq%2BWBuaAsWFTIxEA--&ref=201574f5-816e-4817-a88b-4b095926e817&filter=brand_fa%3AAMD&layout=grid&sort=price_asc

amd llano are getting out of stock now
time to roll

 

[sarinaide]

Well if anything a Trinity is Piledriveresque but it is not a full Piledriver per se, it is difficult to take x and y and come to z conclusion.

The other stand out like a sore thumb aspect is how a low powered Llano still beats Intels creme de la creme fastest chip on the planet in CPU-GPU performance, it is a positive step for AMD and one step closer to true HSA computing.

 

[Regor245]

http://images.anandtech.com/graphs/graph6347/50398.png
thats as per my guesstimations 😗
that
pd is better than k10 in per core (but almost same performance per ghz)
and is almost same to 955/a8 in multithreaded 😀

They are clocked higher.

How about clock for clock? Trinity and Llano.

^
they were like this
pd/trinity will be equal/near to phenom/k10 in performance per ghz
and more clockable thus better per core performance
but due to performace loss in a module , 4ghz module will be equal/near to 3.2ghz 2xk10 cores in multithreaded

and thus 8350 looks like 20-25 % better than 1090t and 8150 in multithreaded
and ~10% better than 1090t in single thread
everything at stock, with marginal error in prediction 😛 😀

How about clock for clock and disabling 2 Cores on FX-8350 vs 1090T?

I think PD has lower IPC than Star Cores.

But i'm sure PD runs cooler and will overclock higher.

___

The FX-4350 will be around 20% percent faster than A10-5800K clock for clock, due to L3 Cache. 😀

Just like Athlon II and Phenom II.

 

[truegenius]

all these are my guesstimations (some guesses and some estimatimations blended together) only
and i am a noob ( New, inexperienced person) 😀

How about clock for clock? Trinity and Llano.

as per me clock for clock (i call it performance per ghz per core 😀 )
it is almost same as of star

but since trinity is clocked higher then it is performing better

How about clock for clock and disabling 2 Cores onFX-8350 vs 1090T?

let performance of pd, bd and k10 be 100, 90 and 105 points per ghz per core
so performance per core of 8150, 8350, 1090t will be 3.6x90, 4x100, 3.2x105 respectively
i.e, 324 , 400, 336 points respectively

now for multi threading
taking work done by 1core as 1 unit
then work done by 2 core of bd, pd and k10 will be 1.55, 1.55 and 1.95 respectively
so performance of 2 cores of bd, pd and k10 (8150, 8350, 1090t turbo off, stock clock) will be 1.55x324, 1.55x400, 1.95x336
i.e, ~500, 620 and 655 points

so 8350 on 6 core will be equal to 3x620=1860 points and 1090t on 6 cores will be equal to 3x655=1965 point

and thats what i call my noobish style 😉
cann't expect brain/inteligent/factual talk from a noob like me 😀

 

[fazers_on_stun]

I agree on that front, but its not like HTT is going to have a huge effect one way or another. Also worth noting is the clockspeed edge the A10 has...

My point being: In CPU benchmarks, Piledriver is still going to trail Ivy by a bit, though it looks like my 8% prediction was just about right.

Well according to those S/A comparisons, P(retty)D(issapointing) will be trailing IB by consideraby more than just "a bit" it seems, although we don't have any actual PD chip reviews out yet. And on an IPC basis, doesn't look like the PD core beats a BD core by more than 5%. Interestingly enough, that 5% improvement matches what IB did compared to SB..

Also looking at the power consumption figures in Chris' test here and Anand's on AT, what happened with the RCM (resonant clock mesh) technology?? The low idle power figures seem more like AMD having finer-grained power gating capability, and shutting off more of the CPU when not in use.

 

[$hawn]

They are clocked higher.

How about clock for clock? Trinity and Llano.

How about clock for clock and disabling 2 Cores on FX-8350 vs 1090T?

___

The FX-4350 will be around 20% percent faster than A10-5800K clock for clock, due to L3 Cache. 😀

Just like Athlon II and Phenom II.

Clock for clock comparison only gives you the IPC difference, which is interesting for people like us, who want to get a glimpse of the architectural differences.

What you are forgetting is the idea behind the BD design. AMD knew that BD cores wud have less IPC, but they tried to adjust for that by providing higher clocks.

What's the point of comparing a A10 at 3 GHz vs a Llano?? Is anyone gonna run their A10 at that speed in real life??

-----

20% extra performance just for L3 cache is a bit too generous.. .for games maybe...but not other stuff....don't generalize

 

[Dangi]

Suppose you went shopping for a new car, maybe a sports car, but there were no reviews comparing different models to each other - all you could get were the basic facts like length, curb weight, engine horsepower and fuel economy. Short of going to 10 different dealers and actually test-driving each model, how would you get the most performance for your $$?

While benchmarks are nowhere near perfect, and YMMV, they remain an effective tool for most people to gauge relative performance so they can make an informed purchase decision according to their wants and needs..

You are right, but this case is the same as if you watch Top Gear reviewing cars, yeah sure a Focus is a great car, but it cannot simply be compared to a Porsche 911 GT3, they don't belong neither the same category not the same price range.

 

[fazers_on_stun]

You are right, but this case is the same as if you watch Top Gear reviewing cars, yeah sure a Focus is a great car, but it cannot simply be compared to a Porsche 911 GT3, they don't belong neither the same category not the same price range.

Heh, you are right of course - my post left out any mention of comparing sports cars "within the same budget range", arguably the most important point .

However, having been here on Tom's for over 10 years now (including before I registered in 2006), I've seen a 180 degree turnaround on CPU benchmarks and power consumption. 10 years ago the AMD fans were touting both benchmarks and power consumption with K8 vs. P4. Now for some odd reason (such as not being competitive maybe), we are being urged to ignore both. While that is what AMD marketing urges with their "our CPU performance is 'good enuff'", this is an enthusiast site and I doubt many here will take said advice. And my guess is that if Haswell starts beating Trinity in GPU benchmarks next spring, we'll be told that we should ignore graphics and gaming benchmarks as well..

What I find puzzling is how loyal and all-forgiving many AMD fans seem to be. If Intel overpromised and underdelivered to the extent that AMD has for the last 5 years (Barcelona and Bulldozer), I'd have dumped them long before now.. So hopefully PD actually turns out to be closer to AMD's promised "15%" improvement, rather than one-third that amount. Otherwise AMD's credibility will drop to around used-car-dealer levels 😀..

 

[de5_Roy]

Well according to those S/A comparisons, P(retty)D(issapointing) will be trailing IB by consideraby more than just "a bit" it seems, although we don't have any actual PD chip reviews out yet. And on an IPC basis, doesn't look like the PD core beats a BD core by more than 5%. Interestingly enough, that 5% improvement matches what IB did compared to SB..

Also looking at the power consumption figures in Chris' test here and Anand's on AT, what happened with the RCM (resonant clock mesh) technology?? The low idle power figures seem more like AMD having finer-grained power gating capability, and shutting off more of the CPU when not in use.

trinity has at least one draw: integrated gpu.
i have asked this before - what kind of attraction does piledriver cpu have? l3 cache? 😗
idle power is not just trinity's credit. imo motherboards and motherboard companies deserve some credit too. overclocked consumption/efficiency seems like the same as bd.
@regor245:
this article has a10 5800k and a8 5600k pitted against an a8 3870k (3.0 ghz) all at stock settings -
http://www.tomshardware.com/reviews/a10-5800k-a8-5600k-trinity-apu,3241.html
here's a single bench with 'fixed' clockrates (compare 3 ghz llano against other cpus with no gpu bottleneck):
http://www.tomshardware.com/reviews/guild-wars-2-performance-benchmark,3268-7.html
keep in mind that it is a single application, it is just one single game thus not a measure of general performance and the editor's jab is 100% justified.

 

[Dangi]

Heh, you are right of course - my post left out any mention of comparing sports cars "within the same budget range", arguably the most important point .

However, having been here on Tom's for over 10 years now (including before I registered in 2006), I've seen a 180 degree turnaround on CPU benchmarks and power consumption. 10 years ago the AMD fans were touting both benchmarks and power consumption with K8 vs. P4. Now for some odd reason (such as not being competitive maybe), we are being urged to ignore both. While that is what AMD marketing urges with their "our CPU performance is 'good enuff'", this is an enthusiast site and I doubt many here will take said advice. And my guess is that if Haswell starts beating Trinity in GPU benchmarks next spring, we'll be told that we should ignore graphics and gaming benchmarks as well..

What I find puzzling is how loyal and all-forgiving many AMD fans seem to be. If Intel overpromised and underdelivered to the extent that AMD has for the last 5 years (Barcelona and Bulldozer), I'd have dumped them long before now.. So hopefully PD actually turns out to be closer to AMD's promised "15%" improvement, rather than one-third that amount. Otherwise AMD's credibility will drop to around used-car-dealer levels 😀..

Yeah AMD is far behind Intel and it's only shoot is HSA, the only thing AMD is better than Intel right now is in iGPU and GPGPU performance, but Intel with his huge wallet can match AMD in 5 years or even catch up earlier with some kind of deal with Nvidia, as Nvidia was looking for a deal with Intel to use theirs fabs.

But AMD did a good thing believing in GPGPU performance, yeah the results have been subpar so far but I don't think Intel has more room left to improve single CPU performance, and moars cores is not the right answer, at least for general purpose.


But amdfanboys and intelfanboys must have high hopes in AMD, if AMD ends up closing Intel will be the only one left, meaning monopoly that is not beneficial for customers 'cause I still can see ARM in the mainstream PC no matter how much they improve their Cortex AXX

 

[davemaster84]

Guys any word on the 8350? 😉

 

[gamerk316]

http://www.anandtech.com/show/6347/amd-a10-5800k-a8-5600k-review-trinity-on-the-desktop-part-2/2

The good news is that single threaded performance is definitely up compared to Llano. Piledriver likely has some to do with this, but so does the fact that the A10 can run at up to 4.2GHz (~4GHz typically) with one of its cores active compared to the 2.9GHz clock speed of the A8-3850. Compared to the Bulldozer based FX-8150 there's a slight (~6%) increase in single threaded performance. Although I don't expect anyone will be cross shopping a Trinity APU and a FX CPU, it's important to keep an eye on progress here as we'll eventually get a high-end quad-module/eight-core Piledriver CPU.

Note that compared to even previous generation, low-end Intel CPUs without turbo there's a huge gap in single threaded performance. If we look at the gap AMD has to make up vs. Ivy Bridge it's not pretty. Intel's Core i3 3220 manages a 27% performance advantage over the A10-5800K. Even if Steamroller is able to deliver a 15% increase in performance at the same clock speed, there will still be a gap. And we're not even talking about how Haswell will grow this gap. For the foreseeable future I don't see AMD closing the single threaded performance gap. Jim Keller's job is to fix this problem, but it'll probably take 2 - 3 years to get there.

More or less as expected.

More worrying:

The multithreaded test shows the other end of reality: in heavily threaded floating point workloads it's possible that we'll see a regression compared to Llano. Remember the Bulldozer/Piledriver architecture prioritizes integer over floating point performance. Truth be told this regression is pretty rare in our tests, but until we get to Steamroller we will still see these types of situations.

As I theorized. AMD has started to sacrifice its threaded performance to try and fix up its per-core deficit, and combined with its lacking FP performance to begin with, you have the possibility of a performance regression.

Moving to threaded Integer benchmarks though:

Here AMD manages a 16% performance advantage over the Core i3 3220. I'd even go as far as to say that Trinity would likely beat any dual-core Intel machine here. The performance advantage is somewhat artificial as Intel purposefully removes turbo from its dual-core desktop CPUs. This should be AMD's best foot forward, but once again it'll likely take Steamroller for this design to start to make sense.

Again, as expected given the architecture.


On the whole, on the CPU side of the house, we see performance close to matching the i3-2100/2330, depending on workload. Only integer heavy threaded benchmarks show Trinity pulling significantly ahead.

 

[sarinaide]

As Anandtech said, its foolish to cross a APU with a full PD or BD module, they are not the same peas in the pot so to take trinity and say that is PD is incorrect because Trinity is a watered down Piledriver module.

 

[esrever]

The threaded performance of the FPU has always been lower on piledriver. This goes back to the tom's review months back. Its 2 FPUs vs 4 and the fact that they are that close is pretty amazing. Considering I don't see a general user using any of the floatpoint heavy software use in the reviews, I don't see a problem. Everything else is better all around.

 

[gamerk316]

Considering I don't see a general user using any of the floatpoint heavy software use in the reviews

Games *cough*

 

[Regor245]

Ok whatever say you guys 😛

thanks

 

[esrever]

Games *cough*

well its given that these APUs aren't ment for high end gaming, they are generally going to be limited to the performance of the integrated and maybe crossfired with a low end discrete. The i3 sandy is only marginally faster compared to the FX4100 when you buy a <$200 gpu so I would think trinity will hold its own even with a 7850 against the ivybridge i3s.