AMD Trinity news (looks, its near to launch?)

[Stukov]

In my years of testing AMD and Intel processors, I have yet to come across a situation like this from either camp. I've seen previews of AMD and NVIDIA GPUs but in every case, there has been a clear determination of the NUMBER of benchmarks allowed rather than cherry picking the ones which make a given architecture look the best.

As per Scott's post, this is about the way it was handled and other sites paying lip service by posting only certain benchmarks in order to make an architecture look better than it really is. This is about information filtering and control of the press. You can't forget that search results on Google, Bing, etc are largely influenced by date / time of posting and content. This means anyone looking for a Trinity article weeks or even months from now will stumble across these articles and have an overly positive representation of the architecture.

We have a very similar process in effect: we refuse to post a reviews at launch unless they can be absolutely, positively up to our readers' expectations. This lack of time is precisely why a Trinity desktop review won't be posted until sometime after launch and why we still don't have a GTX 650 review live. Unless we can completely be confident with the results, I'd rather loose traffic than post something that will steer readers in the wrong direction. I know for a fact that Scott adheres to the same set of ideals.

This isn't to say that Trinity isn't good, because I happen to love AMD's Fusion initiative. Trinity and the new APUs are great candidates for anyone that doesn't give a damn about simple CPU performance. However, there does come a time when pushing certain data ahead of other items doesn't sit well....and this is one of those times.

Well I read Anand's article and the article linked here and I didnt see any mentioning that you couldnt run any game benchmark, right? Or did I miss it? Anand specifically stated he wouldnt have had time to do full CPU benchmark anyway, so it was more like a part 1 and part 2 review (well still closer to preview).

Maybe I'm misunderstanding still, but is this any different than Nvidia (or AMD) allowing sites to benchmark the compute abilities of their chips a week prior to official launch, and on launch its full systems go for testing?

[Russian]

Well I read Anand's article and the article linked here and I didnt see any mentioning that you couldnt run any game benchmark, right? Or did I miss it? Anand specifically stated he wouldnt have had time to do full CPU benchmark anyway, so it was more like a part 1 and part 2 review (well still closer to preview).

Maybe I'm misunderstanding still, but is this any different than Nvidia (or AMD) allowing sites to benchmark the compute abilities of their chips a week prior to official launch, and on launch its full systems go for testing?

This stuff has been done before, it just seems like everyone has a short memory of past launches by both Intel and AMD where there was some sort of control requested of what was posted when...

The truth is, I think all of this stuff has been blown wayyy out of proportion and I feel like there wasn't necessarily anything malicious about how it was done. A few days will not make a difference, especially if the hardware reviewers inform their readers that there is more to come in a few days (especially since most reviewers had no way of getting it all done within the timeframe necessary).

[kl0012]

A few days will not make a difference, especially if the hardware reviewers inform their readers that there is more to come in a few days...

In fact, it can make difference for people, who don't read hardware sites on an ongoing basis. If you are not a tech-freak, and you search google to get fast info - here it may make difference, depends on what article comes first.

[Russian]

In fact, it can make difference for people, who don't read hardware sites on an ongoing basis. If you are not a tech-freak, and you search google to get fast info - here it may make difference, depends on what article comes first.

Correct, but these are not being sold as reviews. They're being sold as previews. People generally base their opinions on Reviews not Previews. If the site is an honest one, they will give full disclosure and it won't be an issue.

[god_43]

....wow, wtf are y'all getting riled up for? a preview? intel does the same ing thing! you know over the past few days, i have seen a lot of anti-amd bs across the entire Internets...this must mean amd is doing something right.


http://www.xbitlabs.com/articles/cpu...hitecture.html


http://www.anandtech.com/show/3871/t...-wins-in-a-row

[kl0012]

Russian
god_43

Guys, this "preview" thing is really boring. Sure, you can always find excuses for inadequate management decisions, but, common, what is the point of doing "architecture previews" for something, which is already few months in the market (mobile version)? Again, saying "Intel does the same..." is not an excuse as well.

[EvilOne]

Again, saying "Intel does the same..." is not an excuse as well.

Intel usually gets away with it, AMD gets shot down every time. AMD needs to realise they can't use the same rules as intel, they dont have enough 'pull' to get away with it.

No disrespect to AMD, that disrespect should go to the review sites that make this possible for Intel.

[kl0012]

Intel usually gets away with it, AMD gets shot down every time. AMD needs to realise they can't use the same rules as intel, they dont have enough 'pull' to get away with it.

No disrespect to AMD, that disrespect should go to the review sites that make this possible for Intel.

To be honest, I did not hear that Intel did something like this before. (Not saying that Intel never did it - I just didn't hear). I know Intel demonstrated some benchmarks here and there in highly controlled environment (Conroe, e.t.c). AMD did it as well. But it is something different this time. I never heard about selective NDA before. Can be a precedent.

[tbone8ty]

http://www.legitreviews.com/article/2043/10/

Preliminary power consumption looks promising!

[Manicdan]

http://www.legitreviews.com/article/2043/10/

Preliminary power consumption looks promising!

a bit of math for ya from that review comparing power consumption and fps for their game test (sleeping dogs)
the A10-5800k used 2% more power than A8-3870k and 18% more than i7-3770k in gaming.
but in fps for their test was 24% more than the A8 and 68% more than the i7
meaning that even in gaming things became much more efficient.

fps/watt = .3 for the A10, .24 for the A8 and .21 for the i7

those numbers could be a result of the gpu portion becoming more efficient, or the cpu portion not needed as much power, but probably both. it would be cool to know how much each section of the APU draws in gaming though.

[tbone8ty]

I think anandtech did that for gpu

[The Stilt]

a bit of math for ya from that review comparing power consumption and fps for their game test (sleeping dogs)
the A10-5800k used 2% more power than A8-3870k and 18% more than i7-3770k in gaming.
but in fps for their test was 24% more than the A8 and 68% more than the i7
meaning that even in gaming things became much more efficient.

fps/watt = .3 for the A10, .24 for the A8 and .21 for the i7

those numbers could be a result of the gpu portion becoming more efficient, or the cpu portion not needed as much power, but probably both. it would be cool to know how much each section of the APU draws in gaming though.

55-60% for the CUs and 40-45% for the UNB (CNB & GNB combined) based on my own measurements.
Please note that on K15h FX-series CPUs the CNB alone was rated to ~14.75W.
On Trinity it should be a bit less.

The measurements were done under maximum FPU & SIMD/RB stress so the numbers most likely do not align up for Gaming.

[Aten-Ra]

Intel Conroe performance previewed
by Scott Wasson — 3:56 PM on March 9, 2006

http://techreport.com/review/9538/in...ance-previewed

The test setup
I should start by saying that we very rarely publish benchmark numbers obtained at trade shows or in other sorts of pre-arranged settings outside of our own test labs. We have elected to do so in this case because of an extraordinary opportunity to get an early glimpse at a brand-new CPU architecture in action. You should know, though, what the test conditions were like.

We used test systems pre-configured by Intel before the show, and we had very limited time to conduct testing or inspect the systems. We were not allowed to look inside of the case of either PC, and the scope of the benchmarks we were allowed to run was defined by Intel. We weren't given the leeway to record our own custom timedemos for the games, and we didn't have enough time to run each test three times or even reboot between the tests. Intel PR reps painted timedemo results on the screen with white-out after our Quake 4 test run.

[rk7p5]

Now that we have all these powerful igpu's, when will we be able to use them for more than games? I want to see Amd's version of quicksync! It would be awesome if people started using these igpu's to speed up apps in general. Most gaming pc's have discrete gpu's and won't use these igpu's, so they just sit there going to waste. It's not like we can choose to buy just cpu's alone anymore, there's very few choices for just plain old cpu's. Also i'm glad amd finally has something better then the competition, let's hope they speed up the cpu parts a bit more.

[Particle]

5800K spotted on eBay

[Maxforces]

Correct, but these are not being sold as reviews. They're being sold as previews. People generally base their opinions on Reviews not Previews. If the site is an honest one, they will give full disclosure and it won't be an issue.

previews you do 1-2-3 months before launch like tomshardware DID. Not during several days. This is review part 1 and part 2

Intel Conroe performance previewed
by Scott Wasson — 3:56 PM on March 9, 2006

http://techreport.com/review/9538/in...ance-previewed

Look at the date of preview and launch March 9 vs 27 July. And now look at AMD preview rather review 5 days difference

[cezar]

First full Trinity review:

http://www.hartware.de/review_1534.html

[Greg83]

Full review !!! lol
that is as bad of a review as the preview.

Single game tested on the built in gpu. rest external gpu
Oh and single resolution done on the built in gpu too. Come on ppl gonna use these with 1080p/720p

[cezar]

Next one:

http://www.overclockers.ru/lab/49852...K_Trinity.html

[Greg83]

there we go. for built in gpu performance, its exactly as expected, smoking intel in the price category
and appears like the preview with Sleeping Dogs, was one of the worst case scenarios at that close to 720p res. what a first image ...

[god_43]

First full Trinity review:

http://www.hartware.de/review_1534.html

...that review makes no sense? why are they benching it against cpus that are double the price?the a series competes against i3-i5 (they have i5).

[cezar]

@god_43

Maybe because it was midnight and they had too much beer

[tbone8ty]

lol trinity gets destroyed it in cpu benchs but who didn't know that already???

at least its a nice step over the 3870k at the same price and one can have fun OC'n, which is great!

and so we wait for steamroller/GCN

[cezar]

One more review and I'm going to sleep - this time is GB mobo but with Trinity onboard:

http://www.back2gaming.com/review/gi...4-motherboard/

[Greg83]

and vs the twice the price 3770k onboard gpu. wins again. nice nice exactly as expected. too bad the preview wasn't so reflective of how good it really is. and i take it those images were showing it beats the gt 620 at 1080p res in all cases too