Mobile Intel Ivy Bridge vs AMD Trinity?

[IntelUser2000]

No wonder Intel is pouring tons of $ into Haswell, IB is behind Llano more than I thought it would be. Seems Trinity may actually get some decent form factors from OEM.

Not in mobile

http://www.anandtech.com/show/5772/mobile-ivy-bridge-and-asus-n56vm-preview/6

Beats 6620G in 9 out of 15 games

http://www.notebookcheck.com/Test-Intel-HD-Graphics-4000.73452.0.html

Beats 6620G in 7 out of 10 games, often significantly

Though 35W/45W Trinity should be able to get 50% faster than Llano(685MHz clocked 384SP VLIW4), going to 17W will need even bigger sacrifice. Their Trinity presentation says it'll be on par with Llano's on 35W, so we may not see a difference in 17W parts even with Trinity.

 

[georgec84]

^ I just saw that article and posted in another thread.

Honestly, the IB GPU looks pretty good.

 

[LOL_Wut_Axel]

well...i don't belive in 56% increse in GPU too... way too high, i bet in ~35%

but for the 29% CPU, i don't see a problem, the clocks are way higher, and piledriver have tons smalls changes over bulldozer

It's just the Bulldozer architecture with some small tweaks. You can't make any big changes to the same architecture in just a year. It's also in the same process node. I just don't see a 30% CPU improvement on average being realistic.

@Lol_wut

I'll bet you'll see bigger jump when compared to what IB got over SB >

As someone else mentioned... Haswell better be good enough, as it has already been redone once. AMD is not kidding around with APU's, they mean business. Kaveri will most likely be another slap in Intel's face... even with Haswell to hand.

On the server side, Intel has AMD beat right now, but i wouldn't be too surprised if i see AMD catching up with Intel with Steamroller at both perf/$ and perf/W

Intel's shens? Well, as part of Anti-trust settlement, Intel can develop exclusive lines with OEM's and exclude AMD from them. Translation, Intel could feed OEM's money, and it is all good, and nobody would bother them if OEM's happen to do a shoddy job with AMD. How's that for tricky little so and so's.

That's yet to been seen, though. Ivy Bridge is a 5-10% improvement in CPU and a 35-40% improvement in GPU. If Trinity is a 10-15% CPU improvement and a 20-25% IGP improvement then it's only "as" good, or worse.

All I know about Haswell as of now is that it should be another 5-10% improvement on the CPU front and a huge improvement in IGP performance, bigger than SB->IB.

It's easy to see how Intel is completely dominating x86 servers and the professional market when you look at the fact they have a big performance/watt advantage and their CPUs have great integer AND floating-point performance. AMD's own, older 12-core Magny Cours is better than the 16-core Interlagos, so right there you can see how they failed.

Anti-trust suit against Intel is old. I'm talking here and now: 2011 and 2012. Also, there's nothing wrong with exclusivity agreements, whether morally or legally. The only thing is it could have repercussions like what AMD experienced with Global Foundries.

Power saved on the CPU front means more power for the GPU. The GPU is also an VLIW4 Southern Island variant so it probably has PowerTune which allows for higher clocks in the same TDP. The 35W A10-4600m for example is a 4 core 2.5ghz/3.2ghz Turbo with 384 shaders and 685 mhz.

http://item.gmarket.co.kr/detailvie...cd=111111111&pos_class_kind=T&search_keyword=

But if they save power on the CPU that means it'll be a tiny CPU upgrade, and we all know mobile Llano is almost within nothing of bottlenecking the relatively tame 6620G IGP because of its very low single-threaded performance. Not really a smart decision because then you won't be able to extract the higher IGP performance. Therefore, they need to balance it out.

 

[386DX]

Not in mobile

http://www.anandtech.com/show/5772/mobile-ivy-bridge-and-asus-n56vm-preview/6

Beats 6620G in 9 out of 15 games

http://www.notebookcheck.com/Test-Intel-HD-Graphics-4000.73452.0.html

Beats 6620G in 7 out of 10 games, often significantly

Though 35W/45W Trinity should be able to get 50% faster than Llano(685MHz clocked 384SP VLIW4), going to 17W will need even bigger sacrifice. Their Trinity presentation says it'll be on par with Llano's on 35W, so we may not see a difference in 17W parts even with Trinity.

Ivy Bridge is going to do much better on the mobile sector then on the desktop. Like I said in my previous post the Intel IGP performance doesn't drop off as much from there lowest end mobile CPU to the desktop CPU while AMD (Llano and pretty safe bet Trinity) drops off pretty significantly as you move from the desktop to the mobile A8, A6 then A4 CPU. This is why like your links shows that mobile IB is about as fast (or slightly faster) then mobile Llano, while the desktop variant is significantly slower. And this is why I'm going to say right now that 17W Ivy is going to be faster then 17W Trinity in CPU and Graphics.

In AMD's presentation they are claiming that 17W Trinity will be on par with 35W Llano. This I can believe but you have to take what AMD says with a grain of salt. When we think 35W Llano we all think A8-3500M, but remember there's also an A4-3300M that is at 35W. I believe this is the 35W CPU that 17W Trinity is going to be on par with. Think about it if the 17W Trinity was on par with the A8 don't you think AMD would say it instead of the "35W Llano". When you market an "inferior" product you always want to make it sound the "best" you can without lying... so you make your statements as vague as possible. A perfect example of this is the Hyundai commercial, where Hyundai claims there car out accelerates a Porsche. Someone hears that and thinks the Hyundai's gotta be fast it out accelerates a 911 Turbo (the car you think when you hear Porsche) but in reality Hyundai is comparing it to the Porsche SUV. You can even take recent quotes from AMD as examples... like the famous Bulldozer "50% faster, with 33% more cores", or "10-15% improvement in performance per watt", or "upto 29% faster CPU".

 

[IntelUser2000]

386,

That's certainly a possible scenario but historically AMD's graphics division has been pretty strong so they may yet exceed what they claim. Graphics claims on the desktop parts were said to be 30% faster but 3DMark06 scores show a 50% gain. Real gains may end up between both but its an example of beating their own promises.

Realistically CPU also affects performance to some degree, and going to dual core will cost Ivy Bridge a bit. Since the top SV Trinity parts can clock at 685MHz while maintaining 384 VLIW4 units, there's still a lot of leeway for downclocking before ending up slower than predecessor's 444MHz VLIW5 400 unit parts. Which I assume would be under 380MHz.

That's because sometime early this year they increased the performance estimates from ~15% to 20-30%, which those clock speeds seem to agree with. They also increased the GPU performance estimates to 50%.l

They never increased the estimates that drastically. You guys are merely confusing the claims made for Ultrathins, standard Laptop, and Desktops. Here's their original claim all over again:

Desktop: 15% CPU, 30% GPU

Standard Laptop: 25% CPU, 50% GPU

Ultrathin: On par with previous generation CPU but with half the TDP

 

[Olikan]

Not in mobile

http://www.anandtech.com/show/5772/mobile-ivy-bridge-and-asus-n56vm-preview/6

Beats 6620G in 9 out of 15 games

ivy -8Gb- DR3-1600
Llano -6Gb- DR3-1333
:sneaky:
but, yeah...very impressive

It's just the Bulldozer architecture with some small tweaks. You can't make any big changes to the same architecture in just a year. It's also in the same process node. I just don't see a 30% CPU improvement on average being realistic.

mmm... yes, and no...it is small changes, but are a page long of small changes...

 

[Fjodor2001]

Regarding Haswell which I'm also considering for an Ultrabook, in the Anandtech Ivy Bridge review the following was stated:

"Internally Haswell is viewed as the solution to the ARM problem. Build a chip that can deliver extremely low idle power, to the point where you can't tell the difference between an ARM tablet running in standby and one with a Haswell inside."

I have some questions regarding that:

1. Approximately what is the idle power consumption of the low power mobile Sandy Bridge / Ivy Bridge CPUs today (i.e. the CPU alone)?

2. Even if the idle power consumption of the CPU hits a theoretical 0 Watts, how much impact will that have on the total battery life of an Ultrabook? I mean the display, HDD/SSD, RAM, WLAN etc will still be consuming power. So approximately how much increased battery life would 0 Watts CPU power consumption translate to (assuming the power consumption of all other parts stay the same)?

3. I know Intel has claimed 10 days "connected standby" for Haswell, see here. But what does that mean? I assume it's when everything is shut off (display, HDD/SDD, etc) and the CPU only automatically and periodically resumes briefly from sleep to check email and so on and then goes to sleep again, right? So for normal and idle usage (as in 2) above) will Haswell really be much of an improvement with regards to battery life?

 

[Olikan]

1. Approximately what is the idle power consumption of the low power mobile Sandy Bridge / Ivy Bridge CPUs today (i.e. the CPU alone)?

2. Even if the idle power consumption of the CPU hits a theoretical 0 Watts, how much impact will that have on the total battery life of an Ultrabook? I mean the display, HDD/SSD, RAM, WLAN etc will still be consuming power. So approximately how much increased battery life would 0 Watts CPU power consumption translate to (assuming the power consumption of all other parts stay the same)?

3. I know Intel has claimed 10 days "connected standby" for Haswell, see here. But what does that mean? I assume it's when everything is shut off (display, HDD/SDD, etc) and the CPU only automatically and periodically resumes briefly from sleep to check email and so on and then goes to sleep again, right? So for normal and idle usage (as in 2) above) will Haswell really be much of an improvement with regards to battery life?

1- around 10 Watts, for the whole chip and south bridge

2-not much, it's the display that use most of the energy

3- it trully seems so, but you never should trust PR material

 

[Fjodor2001]

1- around 10 Watts, for the whole chip and south bridge

2-not much, it's the display that use most of the energy

3- it trully seems so, but you never should trust PR material

1. But how much of those 10 watts is consumed by the south bridge vs the CPU? The south bridge will be consuming power even if the CPU is using 0 Watts (as Intel claims Haswell will be close to).

2. Ok.

3. I'm not sure how to interpret your answer. Do you mean that the 10 days connected standby likely is true, however Haswell will not have much improved battery life when idle or during normal use?

 

[Khato]

1- around 10 Watts, for the whole chip and south bridge

2-not much, it's the display that use most of the energy

3- it trully seems so, but you never should trust PR material

1. It's more like 10 Watts for an entire Ultrabook at idle. Some designs get even lower. Just take a look at various ultrabook reviews on Notebookcheck.net or extrapolate power usage from the idle battery life tests on Anandtech (for example, the 644 minutes of the Asus Zenbook UX31E at idle with its 48Wh battery would require constant power consumption of approximately 4.2W.)

2. Depends on the screen, but usually they're only responsible for around 10W at maximum brightness. Remember, good LED desktop 24" monitors are in the 30W range now, and they have about 3.2x the surface area to illuminate.

3. Yeah, who knows the exact scenario that they're making that claim in. It's certainly possible though, and idle power consumption should see a marked decrease as well thanks to the integration of the VRM.

 

[LOL_Wut_Axel]

1. It's more like 10 Watts for an entire Ultrabook at idle. Some designs get even lower. Just take a look at various ultrabook reviews on Notebookcheck.net or extrapolate power usage from the idle battery life tests on Anandtech (for example, the 644 minutes of the Asus Zenbook UX31E at idle with its 48Wh battery would require constant power consumption of approximately 4.2W.)

2. Depends on the screen, but usually they're only responsible for around 10W at maximum brightness. Remember, good LED desktop 24" monitors are in the 30W range now, and they have about 3.2x the surface area to illuminate.

3. Yeah, who knows the exact scenario that they're making that claim in. It's certainly possible though, and idle power consumption should see a marked decrease as well thanks to the integration of the VRM.

My ThinkPad consumes 4-6W at idle, and it's not an Ultrabook but a 14" laptop with a standard TDP CPU.

 

[Olikan]

anoter leak that "confirms" that trinity have the same ipc of llano

 

[Abwx]

Not in mobile

http://www.anandtech.com/show/5772/mobile-ivy-bridge-and-asus-n56vm-preview/6

Beats 6620G in 9 out of 15 games

http://www.notebookcheck.com/Test-Intel-HD-Graphics-4000.73452.0.html

Beats 6620G in 7 out of 10 games, often significantly

Yes thanks to a voluntarly crippled 6620...

From the pages you quote :

Granted, the A8-3500M/3520M aren't the fastest Llano parts, and the Llano systems we tested were both using DDR3-1333 memory. Give Llano an MX part and faster memory and performance should improve around 20% (5-10% for the RAM, and 10-15% for the CPU).

 

[IntelUser2000]

Yes thanks to a voluntarly crippled 6620...

From the pages you quote :

Yea only for the Anandtech review that has IVB only on par with Llano.

 

[Vesku]

This person seems to have received the A6-4400M and has been benchmarking:

http://semiaccurate.com/forums/showthread.php?t=6402&page=4

"So yeah, about a 20% increase over A4-3300M for CPU.
And also about a 50% increase over A4-3300M for GPU." - vain

http://semiaccurate.com/forums/showpost.php?p=160952&postcount=91

 

[AtenRa]

In DX-11 games Trinity iGPU could be up to 50% faster than Llano iGPU (Batman AC for example). It is easy to see what games will benefit from the Cayman architecture (VLIW4), just see the performance difference HD6970 has over HD5870.

CPU could be up to 30% faster than Llano (higher IPC than BD + higher clocks than Llano)

One more thing,
I would like to see what will happen when you increase the IQ from Medium to High in those Laptop reviews. Im betting Intel iGPU will not be able to produce playable fps. Llano would be able to produce more than 30fps even in DX-11 games and Trinity will even increase over that.

In lower IQ settings the Intel CPU helps more than Llanos anemic CPU performance. Raise the IQ settings to eliminate the CPU bottleneck of STARs cores and let the iGPU do all the work.

 

[Phynaz]

In DX-11 games Trinity iGPU could be up to 50% faster than Llano iGPU (Batman AC for example). It is easy to see what games will benefit from the Cayman architecture (VLIW4), just see the performance difference HD6970 has over HD5870.

CPU could be up to 30% faster than Llano (higher IPC than BD + higher clocks than Llano)

Lot's of could be's in there.

How about I make one up too?

Mobile IVB could be up to 20% faster than Trinity.

There, IVB is now the better CPU.

 

[Abwx]

Lot's of could be's in there.

How about I make one up too?

Mobile IVB could be up to 20% faster than Trinity.

There, IVB is now the better CPU.

Better CPU but far worse GPU......

http://forums.anandtech.com/showpost.php?p=33340263&postcount=28

 

[Khato]

Better CPU but far worse GPU......

http://forums.anandtech.com/showpost.php?p=33340263&postcount=28

Quite amusing. Not only does the i5 2537m have the slowest HD 3000 implementation of the 17W SKUs (okay, I guess there's one i3 that's worse, the i3 2340UE) but given how much lower performance is I'd wager that their test system was using single-channel memory. The normal 17W Sandybridge SKUs score around 1600 in 3dmark Vantage - http://www.notebookcheck.net/Review-Asus-Zenbook-UX21E-Ultrabook.65506.0.html for example

But that does tell us that likely the highest performing 17W SKU is going to be an A6... and Ivybridge gets a bit more than a 1.3x performance boost in graphics at the same TDP as Sandybridge.

 

[AtenRa]

Lot's of could be's in there.

How about I make one up too?

Mobile IVB could be up to 20% faster than Trinity.

There, IVB is now the better CPU.

Since i dont have a Trinity CPU, i can only speculate with the info we already know.

You would hang me upside down by the balls if i would say that Trinity WILL be 50% faster without having a sample or any other information to verify it

 

[Phynaz]

Nah, that would be cruel.

I'd let you stay right side up ()

 

[Zucker2k]

Regarding Haswell which I'm also considering for an Ultrabook, in the Anandtech Ivy Bridge review the following was stated:

"Internally Haswell is viewed as the solution to the ARM problem. Build a chip that can deliver extremely low idle power, to the point where you can't tell the difference between an ARM tablet running in standby and one with a Haswell inside."

I have some questions regarding that:

1. Approximately what is the idle power consumption of the low power mobile Sandy Bridge / Ivy Bridge CPUs today (i.e. the CPU alone)?

2. Even if the idle power consumption of the CPU hits a theoretical 0 Watts, how much impact will that have on the total battery life of an Ultrabook? I mean the display, HDD/SSD, RAM, WLAN etc will still be consuming power. So approximately how much increased battery life would 0 Watts CPU power consumption translate to (assuming the power consumption of all other parts stay the same)?

3. I know Intel has claimed 10 days "connected standby" for Haswell, see here. But what does that mean? I assume it's when everything is shut off (display, HDD/SDD, etc) and the CPU only automatically and periodically resumes briefly from sleep to check email and so on and then goes to sleep again, right? So for normal and idle usage (as in 2) above) will Haswell really be much of an improvement with regards to battery life?

NTV. http://www.anandtech.com/show/5555/intel-at-isscc-12-more-research-into-near-threshold-voltage

 

[OS]

LOL nice catch on the memory speed/size

 

[blackened23]

anoter leak that "confirms" that trinity have the same ipc of llano

WTF? The trinity is A10, not A6, or am I missing something? Also, most of these gaming benchmarks are CPU limited.

 

[Olikan]

WTF? The trinity is A10, not A6, or am I missing something? Also, most of these gaming benchmarks are CPU limited.

all new models of "A" series are trinity based