HP Envy Sleekbok sneak premieres with AMD Trinity

[Gideon]

http://www.nordichardware.com/news/7...d-trinity.html

AMD A-10-4655M APU which has 4 Piledriver cores working at 2,0 GHz, with a turbo frequency which probably is at 2,3 GHz. The graphics circuits is a integrated Radeon HD 7620G that comes with 384 Radeon cores and looks to have a clock frequency at 497 MHz. The AMD A10-4655M gets a TDP value of 25W

IMHO CPU performance is a bit underwhelming. Instead of roughly Llanos performance at half the TDP we get up to 30% worse performance in 70% of the TDP (almost the same performance per watt).

Unfortunately, they didn't do graphics benchmarks, as 497 Mhz for base clock at 25W looks a lot more promising.

As for some reason HP displayed these APUs instead of rumored more powerful 35/17w brethren I'm cautiously becoming pessimistic about Trinity's outlook. As in the worst case scenario these actually will be the chips shipping to most notebooks while the upper end ones will be Press Editions

I hope i'm wrong though ...

[nonameo]

Yeah~~~ at that performance level you might as well just stick with llano.

[ShintaiDK]

So much for Pilediver...

[sm625]

WoW the 25W AMD part gets curbstomped by a 17W intel in Queen and SinJulia to the tune of around 70%. Super Pi is probably the same. Which means most of the tasks that makes a computer feel fast are going to be something like less than half the speed of the intel.

[Olikan]

Originally Posted by Gideon

I hope i'm wrong though ...

LOL there is a llano in the botton...

Look, the 25W trinity is 11% faster than a 35W llano... in photoworxx
(the 25W trinity is 1% faster than the 35W llano in queen)

the perf/watt is >40% here

[Vesku]

Why would they not run some GPU tests, why?

[ShintaiDK]

Lacking drivers perhaps.

[Gikaseixas]

cpu is weak, what a disapointment

[Arzachel]

Only way you could be disappointed by these scores is by having unrealistic expectations. Trinity matches that Llano on FP performance at a smaller thermal envelope while having tons more GPU power. And the turbo should be kicking in due to the benches stressing only the CPU part, so both are at the same clockspeed.

[nyker96]

Originally Posted by ShintaiDK

So much for Pilediver...

ayayayaya!!! tell me this ain't so AMD! another performance regressions!?

before long, they will be at the days of A64s.

[IntelUser2000]

Originally Posted by Gideon

IMHO CPU performance is a bit underwhelming. Instead of roughly Llanos performance at half the TDP we get up to 30% worse performance in 70% of the TDP (almost the same performance per watt).

This is fine, they meant that with 17W parts.

CPU Queen is a branch prediction sensitive benchmark which is why AMD chips are behind significantly. Photoworxx is an integer benchmark and basically represents performance in multi-media like photo editing.

VP8 is a video encoding test based on Google's codec and its a single precision FP test. I can't tell why they are so behind in the SinJulia benchmark, but its an extended precision floating point test. Looks like it may be due to Hyperthreading on the Intel parts.

[frostedflakes]

FPU heavy stuff like wPrime probably won't see much/any performance per watt improvement over Llano. And as IntelUser mentioned, CPU Queen benchmarks branch prediction which was never a strong suit of the BD architecture. These benchmarks pretty much show off the worst aspects of the architecture. Should be some decent gains in integer workloads, though, probably close to the "up to" 30% claims AMD has made.

Those Ivy Bridge quads decimate, though. The people claiming IB is failure crack me up, because they miss the big picture. The performance per watt improvements in IB allow them to clock those 4C/8T mobile parts very high and get some very impressive performance from them.

[Ventanni]

Those Ivy Bridge quads decimate, though.

I think "decimate" is about the perfect word to describe it.

[IntelUser2000]

I think it'll end up mostly like VP8 and Photoworxx. Besides, the 25% increase claim is at same TDP. I doubt the A8 is the highest 35W part. Sure it performs lower per clock, but is clocked way higher than Llano.

Also, Turbo mode isn't as useful on FP. The higher utilization lowers the frequency potential.

[VirtualLarry]

Originally Posted by Ventanni

I think "decimate" is about the perfect word to describe it.

To kill every tenth man, by random lot? Sure about that word?

[Elixer]

I suppose it will be OK if the price is right, but this is still highly disappointing.

[Gideon]

Yeah, it's kind of strange that such benchmarks were chosen, considering the AMD chips only have 2 FP blocks per 4 cores and they don't turbo, these should be some of the least performing benchmarks for AMD. It is as if someone wanted to show weak chips to begin with in the worst possible light. And to not even benchmark the GPU is ridiculous.

The CPU Queen bencmark is still a dissapointment however. AMD promised greatly improved Branch prediction in Bulldozer, after if it failed to deliver, they supposedly improved it for Piledriver but it's STILL only barely faster than a Llano clocked at 1.4 Ghz, despite the 40% clock advatage (and supposedly more agressive turbo).

However this is disturbing as well:

http://www.anandtech.com/show/5825/h...oks-sleekbooks

The AMD-based Envy 6 will be available on June 20, while the rest of the Envy systems are available now.

Really ? These are harvested bargain-bin Trinity parts considering what we should be having at 17W and 35W and even these will only be available in June ?

[KompuKare]

Originally Posted by VirtualLarry

To kill every tenth man, by random lot? Sure about that word?

+1 good to see someone who knows the original meaning of the word...

[frostedflakes]

Was looking at the same benchmarks for Bulldozer, for comparison an FX-8150 gets 31678 in CPU Queen according to the AnandTech review. With half the cores and 56% of the clock speed (2.0GHz vs 3.6GHz) it should get about 8870. A10-4665M gets 11547, so branch prediction is 30% better? Looks like there might be some pretty solid improvements over BD at least.

[Arzachel]

Originally Posted by IntelUser2000

This is fine, they meant that with 17W parts.

CPU Queen is a branch prediction sensitive benchmark which is why AMD chips are behind significantly. Photoworxx is an integer benchmark and basically represents performance in multi-media like photo editing.

VP8 is a video encoding test based on Google's codec and its a single precision FP test. I can't tell why they are so behind in the SinJulia benchmark, but its an extended precision floating point test. Looks like it may be due to Hyperthreading on the Intel parts.

IIRC only the x87 extension supports extended precision which explains why Trinity does so poorly in that bench.

I'd want to see a higher improvement in branch prediction due to the increased pipeline length but everything else seems about right.

For people looking at the bar charts and then shouting doom and gloom:
i7-3610QM is a 45W Ivy quad
i5-2410M and i5-2450M are 35W Sandy dual cores.

[happysmiles]

with the Asus UX21 at around 700-800 pounds and the Sleekbook being 520 pounds (649 euros converted) it doesn't seem to be too bad of a deal.

AMD made it clear they weren't aiming to beat the i7 Ivy bridge. to me they achieved what they set out to do.

[Vesku]

Looks OK for the TDP but really need to see how the GPU compares to current Llano mobile.

[X13]

Yeeeeeeeeeeeah! This is so bad! Oh noeh ... moar benchies! Gimme moar!

Yeah, definitely will buy a DELL with a Trinity processor as soon as it arrives!.. Why? Because i don't run benchies on my computer. I watch movies, i read pdf's, i play old games, i watch free online porn. I play Eve-Online, Gothic 2 and Risen, adventure stuff and when it will come, Path of Exile! Why do i need to throw away 900+ euros for a hot processor that calculates something i will never use very very very fast? You think the processor is the bottleneck? Buy a SSD! Cheers!

[Olikan]

so i heard you want gpu benches...



keep in mind those chips are for embeed market, but the perf. is very close to retail...

"AMD R-464L. This chip packs 4 CPU cores, clocked at 2.3 GHz, and HD 7660G graphics. These specifications, as well as 35 Watt TDP, closely match mobile A10-4600"

http://www.cpu-world.com/news_2012/2...dded_APUs.html

[nonameo]

Originally Posted by Arzachel

For people looking at the bar charts and then shouting doom and gloom:
i7-3610QM is a 45W Ivy quad
i5-2410M and i5-2450M are 35W Sandy dual cores.

Well, I never really expected trinity to compete with sandy or ivy. I was more looking at the comparison to the A6-3400. I was also wondering why they didn't have an A8-3500 or 3510MX in the mix. You'd think that would be a more appropriate comparison.