Maybe later drivers can bring some of the 3DMark magic to games? 
HD 4000@1.15GHz P3132 GPU score
HD 3000@1.10GHz P1410 GPU score
HD 4000@1.15GHz P3132 GPU score
HD 3000@1.10GHz P1410 GPU score
Trinity vs Ivy Bridge
- Thread starter rpg.314
- Start date
caveman-jim
Regular
The awfulness of their CPU's kinda rules them out.
Right, because you can completely ignore the rest of the product line if the top two products are compared. It's turtles all the way down.
No, because serial IPC matters. Llano was about half of SB there.Yes, just like Phenom II was pure crap because Phenom "1" was?
Even if AMD lives up to it's 10% improvement in Piledriver promise, IB will wipe out that difference. AMD needs to gain on Intel's IPC to matter, even though their GPUs are awesome.
rpg, you point is still moot for "budget gaming" computers (for gaming laptops, even more). And, for media consumption.
Actually, for consumers, it's hard to find anything that will make IVB seem noticeably faster than Trinity. If the consumers are smart, they should get an SSD and will get way better "serial performance". (But probably, they are not smart nor willing to spend those money).
Actually, for consumers, it's hard to find anything that will make IVB seem noticeably faster than Trinity. If the consumers are smart, they should get an SSD and will get way better "serial performance". (But probably, they are not smart nor willing to spend those money).
If you want to play games, you would choose some HSA stuff and wait while your movies are encoded. You can transcode via AMD App too, although slower probably
In case not, yeah then Intel's offers are also compelling. Drivers matter. QuickSync plays but a part here.
In case not, yeah then Intel's offers are also compelling. Drivers matter. QuickSync plays but a part here.
Last edited by a moderator:
Anand tested QuickSync 2.0: http://www.anandtech.com/show/5626/ivy-bridge-preview-core-i7-3770k/17
Expreview too, but they had some driver problems: http://en.expreview.com/2012/02/25/...0k-quick-sync-2-0-performance-test/21329.html
It will be interesting how it will compare against AMDs VCE.
Expreview too, but they had some driver problems: http://en.expreview.com/2012/02/25/...0k-quick-sync-2-0-performance-test/21329.html
It will be interesting how it will compare against AMDs VCE.
Oh I see, it's because of the max quality the aforementioned consumers that couldn't have possibly gotten any advantage by purchasing sandy bridge (or ivy bridge) over llano (or trinity) certainly desire and expect.
Obviously no one wants to encode/transcode their videos at much higher speed and a fraction of the power of other products because of subtle differences the vast majority of people wouldn't see in a double blind test. It makes totally sense
Obviously no one wants to encode/transcode their videos at much higher speed and a fraction of the power of other products because of subtle differences the vast majority of people wouldn't see in a double blind test. It makes totally sense
Oh I see, it's because of the max quality the aforementioned consumers that couldn't have possibly gotten any advantage by purchasing sandy bridge (or ivy bridge) over llano (or trinity) certainly desire and expect.
Obviously no one wants to encode/transcode their videos at much higher speed and a fraction of the power of other products because of subtle differences the vast majority of people wouldn't see in a double blind test. It makes totally sense
Yes, one would probably use QuickSync if his/her software supports it, just like one would use AMD APP/STREAM acceleration probably, or in case of Trinity, VCE.
The point was only to show that the CPU isn't all that bad in every scenario
--
Aside from that, why exactly is Trinity VLIW4? Sure, VLIW4 was ready earlier than GCN, that's a given, but Trinity still features VCE which was introduced with GCN, as well as DX11.1 support apparently, which wasn't part of the only other VLIW4 chip either.
Does it? Honest question, but this is the first time I hear it.
I wasn't speaking for others. I was speaking for my needs.
IPC does not matter anything, total performance (clock speed*ipc) matters.
And no, it was 60-90% of the single-thread performance of SB which had 6% higher clock speed, usually like 70%.
So this means it's serial IPC is about 65-94% if SB's IPC, usually around 75%.
You are now messing bulldozer and llano; Llano is K10, that 10% projected ipc increase is over bulldozer.
So, what is realistic to except:
* About similar IPC than Llano (bulldozer had slightly worse average ipc than K10)
* Much higher clock speeds , on desktop from 3 GHz to 4.2/3.8 GHz.
So, compared to llano, on desktop trinity gives about 40% increase in serial performance, 25% increase in 2-thread performance but almost no increase in 4-thread performance(with 4 cores active shared L1I, L2, decode, fpu eat ipc).
On mobile side the clock rate difference between llano and trinity might be smaller, but still considerable.
Ivy bridge will still be much faster, but AMD is narrowing the gap in single-thread performance.
Last edited by a moderator:
Seriously?
Sources?
40% single thread might be a bit optimistic but it'll be 20% on clocks alone. 4.2 GHz / 3.0 Ghz * 0.85 = 1.19 on clock speed improvement over Llano.
If you assume only 5% IPC gain from BD to Piledriver (removing the L3 would still be in line with Stars > Llano) we should expect a minimum of 25-26% improvement in single threads.
If you assume only 5% IPC gain from BD to Piledriver (removing the L3 would still be in line with Stars > Llano) we should expect a minimum of 25-26% improvement in single threads.
I suspected as much. That's why I only replied to your second post on the matter, not the first. However, thet way you said it did not really implied that your were speaking for yourself. Rather, it sounded like stating something general (at least for me it sounded that way).
Seriously?
Actually, the AMD material on the matter has been worryingly vague. They promise that 10-15 % of yearly performance increase for each new core. But nowhere is it clearly stated that those improvements refer to IPC alone or they do factor in the clock increase.