Trinity review

[Schmide]

I'm sure it is doing more work in prime95, but what about furmark frames? Do you have any data on that?

Furmark is pretty flat? 3-5w between the big contenders A10 vs i7 hd4000.

What is really interesting is the way the graphics turbo can be seen, Prime and Prime + Furmark following the same line on the i7s.

At any rate, my point is simply that it is in an entirely different TDP class, not that it is any faster or slower.

and? That point would be moot in that graph. If you look, the activities that any user would do for prolonged periods, (fifa, web, utube) would actually favor the i7.

 

[Vesku]

Intel may not have altered their public unit pricing but it just wouldn't be possible to have the SB lower end notebooks be priced what they started selling for after Llano if the OEMs were paying the public unit price to Intel. Unless you are trying to say HP, Dell and crew actually took a loss to help Intel?

You are also arguing that Intel spent the time and money to SKU out a higher iGPU boosting product for no other reason than they decided it would be nice to spend money to do so? They have to test and qualify parts, it doesn't make sense to qualify a part for 10ish % more iGPU turbo with no additional change in the cores unless they thought there was a market reason to do so.

That doesn't explain anything. It could easily be because of improved yields and lower leakage.

And Intel didn't "aggressively drop prices" when Llano was launched. Unit pricing didn't change one bit from when it was launched to when after Llano was launched.

Nice try, but argument not found.

 

[Phynaz]

Wow that is good news about VCE. File size is far far more important than encoding time for me. I really dont care how long I have to wait for a transcode... as long as I can set up batches and let them run i really dont care. But I want the smallest possible file size! If quicksync cannot produce that then it is useless.

Then why would you use VCE?

Just encode on the CPU and use any bitrate you want.

 

[Phynaz]

Hence my answer that an i3 would be no match..

I guess we will know this summer.

 

[Olikan]

Then why would you use VCE?

Just encode on the CPU and use any bitrate you want.

were do you found VCE bit-rate info?

 

[Phynaz]

72 watts sounds like desktop territory to me. There is no way that can possibly be acceptable in a modern 14-15" notebook chassis.

You're right, you can't put 72 watts in a notebook chassis.

Therefore that graph is BS.

Funny how people get so desperate to search out information to support their viewpoint that they never consider its source or if it's valid.

 

[Phynaz]

were do you found VCE bit-rate info?

Sounds like you have never done video encoding before.

Every app I've seen allows you to set the bitrate that you want to encode at.

 

[Khato]

Well this should put the trolls to bed:

http://www.notebookcheck.net/typo3temp/pics/ef40e4e968.png

Here we see that the ivy bridge quad consumes considerably more power than the trinity. If you subtract about 8 watts baseline idle power from each you get the ivb consuming over 60% more power at full load! Sure it might be equal in terms of performance, but it does so by running in completely different TDP class. 72 watts sounds like desktop territory to me. There is no way that can possibly be acceptable in a modern 14-15" notebook chassis. It is going to run into cooling problems and probably be throttling after a few months collecting dust in a typical household.

Or we can say that on the one actual application benchmark in those charts, Fifa 12, the A10-4600M drew 49W at the power adapter compared to 39.5W for the i7-3720QM. If you want to again subtract an 8 watt baseline (I don't believe that's a realistic value, but since it's the value you used) then you have the A10-4600M using 30% more power for, well, we don't know what kind of performance difference since they didn't specify what settings were used for the power comparison. The HD 4000 wins at low and medium settings while falling behind at high and ultra.

As for the synthetics... yeah, no question that AMD wins on furmark, and most likely on all graphics workloads. However the ~66W of the i7-3720QM on prime95 delivers more than double the performance of the 40W of the A10-4600M. Further, there is a potential flaw in these power figures - sure they eliminated the LCD from the equation, but they made no mention of eliminating the AC adapter. High-end AC adapters, like those used by power misers with picoPSUs, are at around 96% efficiency, whereas a typical one will be in the 80-85% range peak. Even if there's just a difference of 95% vs 85% efficiency (the i7-3720QM was in an actual production model, whereas the A10-4600M is a reference platform that they want to look good, so chances are quite good they wouldn't have skimped on AC adapter on the off chance it'd be used to measure power consumption) then peak numbers shift to 46.5W vs 61.2W.

 

[Olikan]

Sounds like you have never done video encoding before.

Every app I've seen allows you to set the bitrate that you want to encode at.

atleast in AMD slides, VCE is very close to CPU....
So i thought that it would be very flexiable to bit-rate ratio...

since, we know almost nothing about VCE i truly thought that you found an obscure review XD

 

[Lonbjerg]

some first trinity laptops spoted...
A6-4400 APU is paired with a discrete HD 7520G....446 EUR

http://news.softpedia.com/news/HP-Pavilion-G6-with-AMD-Trinity-Selling-for-570-USD-271232.shtml

LOL...so that APU wasn't cutting it alone but need a GPU to hold it's hand?

 

[LOL_Wut_Axel]

Intel may not have altered their public unit pricing but it just wouldn't be possible to have the SB lower end notebooks be priced what they started selling for after Llano if the OEMs were paying the public unit price to Intel. Unless you are trying to say HP, Dell and crew actually took a loss to help Intel?

You are also arguing that Intel spent the time and money to SKU out a higher iGPU boosting product for no other reason than they decided it would be nice to spend money to do so? They have to test and qualify parts, it doesn't make sense to qualify a part for 10ish % more iGPU turbo with no additional change in the cores unless they thought there was a market reason to do so.

You still haven't clarified regarding what I asked you about.

I don't care about your conspiracy theories about what Intel is doing with X and X OEM. What they pay for the CPUs is widely available information, and part of the reason OEMs were able to gradually lower pricing is that as volume had gone up and chassis manufacturing pricing had gone down, manufacturers were able to keep the same margins and before while lowering pricing.

You act like if Intel was doing this secret thing when they've been giving free bumps to their line forever. The same thing happened with Conroe and Penryn where new steppings came in, testing was done, and better samples with lower leakage and higher clock speeds were released. Same thing happened with Nehalem and Westmere, where as time passed better products were introduced at the same price (for example, i7-920 replaced with 930 and 950; i5-750 replaced with 760). Same thing happened with Sandy Bridge, where newer versions replaced the old (i7-2700K replaced 2600K; i5-2550K replaced 2500K and 2xx0 versions were replaced with 2xxx5 versions). Yes, it is completely worthy of Intel's time to replace products with improved versions at the same price because it keeps consumer interest up and higher sales more than make up for the additional testing and validation.

Now stop acting like if all of this was some Intel secret. It's standard, industry-wide practice, and you've yet to prove it's not (and you won't) because history will show you this has been done for many years.

 

[Vesku]

Not sure how to respond to someone who thinks big OEMs pay list prices...

 

[AtenRa]

On the flip side of the coin, if you cherry pick in favor of the HD 7660G, it's 88% faster in a game almost no one plays: Civ V.

Both Skyrim and Battlefield 3 are FAR more popular PC titles than Civ V.

Im glad you reevaluated your initial statement, yes Skyrim and BF3 are more popular but people also playing Civ V

 

[Lonbjerg]

Why does that APU need a compaion GPU?
That laptop just confirms my point

 

[sm625]

You're right, you can't put 72 watts in a notebook chassis.

Therefore that graph is BS.

Funny how people get so desperate to search out information to support their viewpoint that they never consider its source or if it's valid.

So you are saying this article is bs? http://www.notebookcheck.net/Trinity-in-Review-AMD-A10-4600M-APU.74852.0.html

You are saying they did all their i7 notebook power draw tests completely wrong?

 

[Abwx]

Why does that APU need a compaion GPU?
That laptop just confirms my point

You real point is that you re just showing how ridiculous you are.

HD7520G is a cut down version of low end Trinity s INTEGRATED GPU....

So there s no discrete GPU in this LT.

Edit : there a dual switchable gfx version...

AMD Trinity Quad Core A8-4500M (2.1 GHz)
AMD Radeon HD7640G intégrée à l’APU + Radeon HD7670M 1 Go DDR3 dédié

http://www.laptopspirit.fr/110357/h...trinity-a8-quad-core-6-go-hd7670m-usb-30.html

 

[Phynaz]

So you are saying this article is bs? http://www.notebookcheck.net/Trinity-in-Review-AMD-A10-4600M-APU.74852.0.html

You are saying they did all their i7 notebook power draw tests completely wrong?

I didn't say any of those things, did I?

 

[Khato]

So you are saying this article is bs? http://www.notebookcheck.net/Trinity-in-Review-AMD-A10-4600M-APU.74852.0.html

You are saying they did all their i7 notebook power draw tests completely wrong?

A 72W power draw by the AC adapter does not equate to 72W being used by the notebook. As I stated in a prior post, they made no indication as to whether a common AC adapter was used across the tested platforms. If they're using the AC adapters that came with the respective systems, then their results will be skewed accordingly.

 

[Schmide]

You're right, you can't put 72 watts in a notebook chassis.

Edit: Notebook class? I didn't know there was a limit?

They do call it a notebook here.

Edit: http://hothardware.com/Reviews/Alienware-M18x-Gaming-Notebook-A-Tale-of-Two-GPUs/?page=8

Finally, with respect to power consumption, under heavy gaming testes, the dual Radeon HD 6990M powered machine pulled about 220 Watts at its peak consumption. The M18x, configured with a pair of GeForce GTX 580M GPUs in SLI, pulled about 240 Watts at its peak draw. Both machines spiked an additional 60+ Watts if the machine was under a game load and also charging the battery.

 

[lau808]

is there a reason no reviews went more in depth like for example comparing ipc vs llano, phenom 2, bulldozer, 2500k/3750k?

 

[LOL_Wut_Axel]

http://www.tomshardware.com/news/amd-intel-cpu-apu-processors,15741.html

That goes squarely against the reasoning that Intel is "afraid" of AMD. AMD doesn't want to make fast processors anymore, meaning they'll keep pushing their "good enough" mantra. Let's see for how long that lasts them.

Edit:

This also confirms what I said earlier about the Bulldozer architecture being fundamentally broken and therefore not allowing AMD to be competitive with Intel in CPU performance.

 

[Homeles]

This also confirms what I said earlier about the Bulldozer architecture being fundamentally broken and therefore not allowing AMD to be competitive with Intel in CPU performance.

The Bulldozer concept is quite sound. The first iteration was awful, but Piledriver's a great step forward.

 

[Soulkeeper]

I actually prefer the modular design over hyperthreading.
Problem is Intel leads in so many areas AMD made their otherwise good idea look bad.
Things like cache latency/bandwidth, intel's turbocore is impressive, process tech, and various minor optimizations for power usage.

 

[Olikan]

hehe...the leaks started...

 

[LOL_Wut_Axel]

The Bulldozer concept is quite sound. The first iteration was awful, but Piledriver's a great step forward.

Not sure if serious... unless you found Netburst to be "sound". If Bulldozer was sound, AMD wouldn't be downplaying the negative of it having underwhelming performance. Piledriver is rumored to be 10% faster than Bulldozer and have 10-25% power savings.

That's really barely better than what Intel did with Ivy Bridge, which was an increase in performance of 5% and power savings by the same percentage as Piledriver.

http://www.techspot.com/news/48742-rumor-amd-piledriver-fx-cpu-production-to-begin-q3-2012.html

Of course, that's assuming the info is accurate, which it doesn't look not to be since it's only a revised Bulldozer on the same process node.

And while the modular design sounds good for desktops theoretically, in practice Hyper-Threading is a better solution, especially if you have high single-threaded performance (which AMD does not have).