Trinity review

[Skurge]

All games on IGP are pushing it...but not enough.
It's a step backwards...and why hamper the CPU (TDP)...with IGP performnace?

Give me a clean CPU core...I'll add a to the system GPU if I need it...but I have no use for a IGP anymore than I have a use for a console.

Thats what Sandy Bridge-E and AMD FX (although FX is a non starter, I had to inclde it.) So you coming into this thread complaining is really pointless as the IGP in these CPU doesn't hamper them.

SB-E with all that cache and memory bandwidth can't even pull away from a 2700K/3770K. I'm talking about the 3820 as it is also a quad core.

 

[Lonbjerg]

Thats what Sandy Bridge-E and AMD FX (although FX is a non starter, I had to inclde it.) So you coming into this thread complaining is really pointless as the IGP in these CPU doesn't hamper them.

SB-E with all that cache and memory bandwidth can't even pull away from a 2700K/370K7. I'm talking about the 3820 as it is also a quad core.

Try comparing with IV-E.
Then come back and share your results.

Failed comprison is failed comparision.

 

[Skurge]

Try comparing with IV-E.
Then come back and share your results.

Failed comprison is failed comparision.

Well IB-E isn't here. You have yet to show me how IGPs are handicapping CPUs. SB-E is DOUBLE the die size of SB and only shows a tangible advantage in the few multithreaded apps on desktop. If you are a server or HPC user, that you could say that they might be, but on the client side they aren't. In fact with things like quicksync and handbrake. The "gimped" CPUs with IGPs outperform the multicore monsters with 3x the cache and double the cores.

 

[Lonbjerg]

Well IB-E isn't here. You have yet to show me how IGPs are handicapping CPUs. SB-E is DOUBLE the die size of SB and only shows a tangible advantage in the few multithreaded apps on desktop. If you are a server or HPC user, that you could say that they might be, but on the client side they aren't. In fact with things like quicksync and handbrake. The "gimped" CPUs with IGPs outperform the multicore monsters with 3x the cache and double the cores.

So you don't understand the concept of "A hammer is for nails" and "A screwdriver is for screws"...gotcha.

 

[Phynaz]

IB IGP has been overhyped....

Let s see when either a 45W Trinity is launched or a 35W IB
is used instead of the 45W or sometimes 55W parts...

Funny how you make power consumption an issue now. Why wasn't it an issue for all your Derpdozer posts?

 

[Olikan]

Funny how you make power consumption an issue now. Why wasn't it an issue for all your Derpdozer posts?

because we are talking about mobile here :twisted:

 

[HexiumVII]

So is Trinity at least up on par with Nehalem?

 

[Abwx]

Funny how you make power consumption an issue now. Why wasn't it an issue for all your Derpdozer posts?

That s dangeourous to answer with automatisms ,
one will easily lose any credibility in such case ,
moreover with false assumptions...

Anyway , we are talking about products that have
precise market segmentation , i.e , mobile plateforms
with 17W , 25W , 35 and eventually 45 and 55W
thermal envelloppes.

Going from 35W to 45 allow at least 20% better perfs
for GPU+CPU or as much as 40% for either the CPU
or the GPU if they benefit exclusively from the added TDP.

 

[LOL_Wut_Axel]

So is Trinity at least up on par with Nehalem?

No.

 

[pelov]

So is Trinity at least up on par with Nehalem?

It's pretty close depending on which mobile Nehalem chip you're talking about. Single-threaded performance is still behind, maybe 5-10%, but multi-threaded performance the Trinity chips do better -- which means the old Llano chips performed better as well because AMD hasn't pushed the multi-threaded performance much (or at all) due to CMT tax that Llano didn't have.

http://www.notebookcheck.net/Intel-Core-i3-350M-Notebook-Processor.23752.0.html
http://www.notebookcheck.net/Intel-Core-i5-430M-Notebook-Processor.23750.0.html
http://www.notebookcheck.net/AMD-A-Series-A10-4600M-Notebook-Processor.74065.0.html

Given the small module size and 50% die space restricted to GPU, it's pretty good when compared to Nehalem. Considering that and perf-per-watt/battery life it makes it a no-brainer when those 2 are compared.

clarkdale

It looks like it depends on the models. The SB i3's trade blows with the A10 APU on some benchmarks but the i5's seem to consistently beat it out. It's sort of messy, really. Even some of the old Phenom II's seem to perform competitively in some of those benchmarks along with the C2Duos.

 

[Phynaz]

Going from 35W to 45 allow at least 20% better perfs
for GPU+CPU or as much as 40% for either the CPU
or the GPU if they benefit exclusively from the added TDP.

Ok, so power consumption doesn't matter on desktop, but does in mobile. Gotcha.

Got anything to back up your other statements? If it was as easy as that, why would AMD leave so much performance laying on the floor?

 

[LOL_Wut_Axel]

]It's pretty close depending on which mobile Nehalem chip you're talking about. [/B]Single-threaded performance is still behind, maybe 5-10%, but multi-threaded performance the Trinity chips do better -- which means the old Llano chips performed better as well because AMD hasn't pushed the multi-threaded performance much (or at all) due to CMT tax that Llano didn't have.

http://www.notebookcheck.net/Intel-Core-i3-350M-Notebook-Processor.23752.0.html
http://www.notebookcheck.net/Intel-Core-i5-430M-Notebook-Processor.23750.0.html
http://www.notebookcheck.net/AMD-A-Series-A10-4600M-Notebook-Processor.74065.0.html

Given the small module size and 50% die space restricted to GPU, it's pretty good when compared to Nehalem. Considering that and perf-per-watt/battery life it makes it a no-brainer when those 2 are compared.

clarkdale

It looks like it depends on the models. The SB i3's trade blows with the A10 APU on some benchmarks but the i5's seem to consistently beat it out. It's sort of messy, really. Even some of the old Phenom II's seem to perform competitively in some of those benchmarks along with the C2Duos.

Very easy to do this comparison when you compare the highest-end A10 APU against the low-end Clarkdale Core i3. No, it's slower when you compare overall results and don't nitpick to favor your point of view.

And the SB i3 is flat-out faster than any A10 CPU.

 

[ElFenix]

So you don't understand the concept of "A hammer is for nails" and "A screwdriver is for screws"...gotcha.

so basically:
1. make a claim
2. when presented evidence that the claim is false, refer to unreleased processor
3. when it's pointed out that the processor is unreleased (and, frankly, not likely to change much), fall back on platitudes.

 

[pelov]

And the SB i3 is flat-out faster than any A10 APU.

Is this another one of the famous LOLWTFAXELBBQ predictions that falls flat on its face as soon as someone uses actual benchmarks/statistics to prove you wrong?

Deja vu?

The SB i3 doesn't kick Trinity's ass in all CPU benchmarks. Not even close. Trinity performs quite well against the i3 SB's, losing some single-threaded benchmarks (and not even all of them, btw) and winning most multi-threaded benchmarks. What's holding the SB i3 back is the lack of turbo which you can clearly see in the results from the i5 SB which does indeed kick Trinity's ass in CPU benchmarks. So you're wrong... again.

 

[Lonbjerg]

so basically:
1. make a claim
2. when presented evidence that the claim is false, refer to unreleased processor
3. when it's pointed out that the processor is unreleased (and, frankly, not likely to change much), fall back on platitudes.

No, he cheerypicks the "bar" (avoiiding heavy multitasking senarios, the reason why you would go scoket 1336/2011 in the first place)...and tries to make it sound like it a valid comparsion between last generations workstation CPU vs last generation desktop CPU..talk about comparing bananas to chickens.

His claim cannot be verified or falcified until IVB-E is tested.

But thanks for trying

 

[LOL_Wut_Axel]

Is this another one of the famous LOLWTFAXELBBQ predictions that falls flat on its face as soon as someone uses actual benchmarks/statistics to prove you wrong?

Deja vu?

The SB i3 doesn't kick Trinity's ass in all CPU benchmarks. Not even close. Trinity performs quite well against the i3 SB's, losing some single-threaded benchmarks (and not even all of them, btw) and winning most multi-threaded benchmarks. What's holding the SB i3 back is the lack of turbo which you can clearly see in the results from the i5 SB which does indeed kick Trinity's ass in CPU benchmarks. So you're wrong... again.

LMAO at you saying the A10 APU is faster or comparable by pointing at benchmarks which have the ULV i3-2367M which runs at 1.4GHz. Pretty hilarious, especially since you saying this is bad for your argument because you're essentially admitting that the A10 is only as good as a 17W, low-end ULV Core i3 processor.

Just FYI, the 4600M "competes" with 35W Core i3 and i5. And the lowest-end 35W SB Core i3 runs at 2.1GHz, 50% higher than the ULV Core i3 in the review.

I also like how you only post benchmarks for heavily multi-threaded programs.

 

[IntelUser2000]

The SB i3 doesn't kick Trinity's ass in all CPU benchmarks.

-That's a 17W Core i3
-There's only one single threaded benchmark, which is Cinebench

Ok, so power consumption doesn't matter on desktop, but does in mobile. Gotcha.

Actually, Arrandale is worse not just because of TDP, but idle power. Bulldozer proved bad when they tried to scale for absolute performance, but from my memory it wasn't bad when looking at lower chips like the FX-4100.

 

[ElFenix]

No, he cheerypicks the "bar" (avoiiding heavy multitasking senarios, the reason why you would go scoket 1336/2011 in the first place)...and tries to make it sound like it a valid comparsion between last generations workstation CPU vs last generation desktop CPU..talk about comparing bananas to chickens.

His claim cannot be verified or falcified until IVB-E is tested.

But thanks for trying

your claim was that graphics bits were holding back CPU bits. why do we need IB-E to test that? we have SB-E. it's not faster except in a very few situations. SB isn't being held back by the graphics part.

especially since most of the diff between SB and IB is in the graphics, what's the diff between IB-E and SB-E?

 

[pelov]

Or you can just click the links above with the i3 SB's at 35W.
http://www.notebookcheck.net/Intel-Core-i3-2310M-Notebook-Processor.45318.0.html
Usually within single digits while the i3 pulls away in single-threaded. The i5 pulls away in both due to turbo.

None of the reviewers bothered to benchmark Trinity against SB i3's and most only pulled out the i7 IB quads so it's difficult to get a good comparison.

 

[LOL_Wut_Axel]

Or you can just click the links above with the i3 SB's at 35W.
http://www.notebookcheck.net/Intel-Core-i3-2310M-Notebook-Processor.45318.0.html
Usually within single digits while the i3 pulls away in single-threaded. The i5 pulls away in both due to turbo.

None of the reviewers bothered to benchmark Trinity against SB i3's and most only pulled out the i7 IB quads so it's difficult to get a good comparison.

Or you could just be honest and admit you were wrong? Yes, the 35W A10 is faster than a 17W i3. But what you're doing is comparing a low-end ULV processor to AMD's highest-end standard voltage processor.

Comparing the 35W SB i3 and the 35W A10, the i3 is clearly faster overall. Let's not forget either that this comparison isn't all to clear, either. The IB i3 will go against the Trinity A8 for the $500 range laptop market. When looking at CPU performance alone only Core i3+dGPU (HD 7670M) laptops will be going against the A10 for the $600 range. At that point the Core i5 comes into the picture, too.

 

[Schmide]

If you do all the math it would seem that Trinity is about 10% less than a Phenom II IPC. At best you could maybe put it on par if it had the same cache support. Regardless it seems to be 15-20% higher than bulldozer. If that means something.

AMD needs the fix their freaking cache and memory systems now!!!

 

[Skurge]

No, he cheerypicks the "bar" (avoiiding heavy multitasking senarios, the reason why you would go scoket 1336/2011 in the first place)...and tries to make it sound like it a valid comparsion between last generations workstation CPU vs last generation desktop CPU..talk about comparing bananas to chickens.

His claim cannot be verified or falcified until IVB-E is tested.

But thanks for trying

http://www.anandtech.com/show/5276/intel-core-i7-3820-review-285-quadcore-sandy-bridge-e/2

3820 vs 2600K. Same Die size, same number of cores (2600K is clocked 200mhz lower) the 2600K has an IGP, the the 3820 has 2MB more cache and double the bandwidth.

Performance is EXACTLY the same. Power use is EXACTLY the same. Actually, the 2600K wins cause it has quicksync.

If you want to compare the 3930K, it's double the Die size. Give the 2600K double the Die size and and you add 2 more cores and the keep the IGP. Argue, the merits of an IGP if you want, but you are showing irrational hatred towards IGPs when the CPU with the IGP is actually a better option than the one without.

 

[ElFenix]

http://www.anandtech.com/show/5276/intel-core-i7-3820-review-285-quadcore-sandy-bridge-e/2

3820 vs 2600K. Same Die size, same number of cores (2600K is clocked 200mhz lower) the 2600K has an IGP, the the 3820 has 2MB more cache and double the bandwidth.

actually the 3820 is much bigger with ~275 million more xtors and 80 sq mm more silicon

 

[pelov]

Or you could just be honest and admit you were wrong. Yes, the 35W A10 is faster than a 17W i3. But what you're doing is comparing a low-end ULV processor to AMD's highest-end standard voltage processor.

That's the same thing the AMD-nuts have been saying about all of the reviews pitting the A10 against the 45W i7 IB and even the 55W i7 IB. It just goes to show you how shitty and half-assed the reviews have been, Anandtech included.

The i3 35W SB doesn't stomp on the A10 either and the "win" isn't as clear (flat-out faster) as you made it seem. It can get pretty close and you can see that.

Let's not forget either that this comparison isn't all to clear, either. The Ivy Bridge Core i3 will go against the Trinity A8 for the $500 range laptop market. When looking at CPU performance alone only Core i3+dGPU (HD 7670M) laptops will be going against the A10 for the $600 range. At that point the Core i5 comes into the picture, too.

That I agree with 100%, but atm we're discussing hypotheticals. When comparing laptop builds there's a lot that goes into it and not just cinebench scores. When you buy a laptop you're buying the entire package and not just the CPU so it's hard to compare competing processors fairly. I think it's much easier to do that when you've got the laptops in hand and you can consider things like price, battery life, discrete GPU vs APU, etc., otherwise we're discussing hypotheticals. We don't know what OEMs will do to Trinity yet. HP has a "sleekbook" for ~$600 with the A10 which is pretty good but we don't know the exact specs/options. The i3 IB's don't even have a release date other than a rough estimation so we don't know how they'll be featured by OEMs and what the prices/options will be.

I think the only certainty is that SB/Llanos will be the best bang-for-your-buck. Some SB i5's with discrete GPUs can be had for ~$600 and Llano crossfired with dGPUs can be had for ~$500. If those get even lower they'll be an absolute steal.

 

[Schmide]

The real test for me will be if AMD actually puts out a high end notebook paired with a turks 6630M or so in Hybrid Crossfire. If they can get it to preform anywhere near Llano, it could be a beast of a gaming book and possible keep up with higher end dedicated graphics at a decent cost.

http://www.techwarelabs.com/amd-llano-a8-3850-apu/4/