AMD CPU speculation... and expert conjecture

[gamerk316]

mantle is 45% faster.

On a specific title, one a specific hardware config, on a specific OS. Could probably get a lot more if they removed the entire DX backend. Could probably get a lot less if they spent less time implementing it.

 

[etayorius]

mantle is 45% faster.

http://www.youtube.com/watch?v=yfXI1pn5los


So if the 7850k uses a 7750, with a 45% increase what would that make it ?

Probably between the HD7850 and HD7870.

 

[Comgen]

mantle is 45% faster.

http://www.youtube.com/watch?v=yfXI1pn5los


So if the 7850k uses a 7750, with a 45% increase what would that make it ?

If this same boost is applied to the APU's, will be an impressive result.
although it would be interesting to know what hardware they were using in this video.

Also I'm curious to see how the 7850k will behave in OC, if good, coupled with some badass high speed ram, it will be a nice toy.
I'll wait the first real&serious tests to decide though.

 

[Cazalan]

Already know from where AMD got the 20% IPC for Steamroller claimed this week-end. It is from using 2nd pass of the x264 benchmark. Kaveri scored 8.53 and Richland 7.09. Both clocked at 3.7GHz under W8.1 and 2133 memory.

Taking this review as base

Kaveri would score ~9.5, which is at the i5-2400 level and only ~8% behind FX-6300 and i5-2500k. It looks very close to my prediction ;-)

That bench doesn't scale linearly. As you see the A10-6700 is 7.89, and A10-6800K is 8.26. That's only a 4.7% gain (8.26/7.89) for a 10.8% clock speed gain (4.1/3.7). Or 4.35% for a 10% speed gain. Multiply by 2 for the 20% gain AMD stated, 8.7%.

So Kaveri @ 3.7Ghz would be 7.89 x 1.087 = 8.58 which is darn close to the 8.53 number by AMD. No need to add a further percentage to the number AMD provided.

As many suspected the IPC gains while great are being hampered by the clock speed reduction for traditional workloads. The GPU gains with GCN are where the big gains are to be found.

 

[noob2222]

So basically kaveri will finally allow an APU to reach the phenom II x4 level of performance.

 

[Cazalan]

So basically kaveri will finally allow an APU to reach the phenom II x4 level of performance.

Well that's using the CPU to do the encoding. An APU would be using the GPU which should be much faster.

 

[juanrga]

Already know from where AMD got the 20% IPC for Steamroller claimed this week-end. It is from using 2nd pass of the x264 benchmark. Kaveri scored 8.53 and Richland 7.09. Both clocked at 3.7GHz under W8.1 and 2133 memory.

Taking this review as base

Kaveri would score ~9.5, which is at the i5-2400 level and only ~8% behind FX-6300 and i5-2500k. It looks very close to my prediction ;-)

That bench doesn't scale linearly. As you see the A10-6700 is 7.89, and A10-6800K is 8.26. That's only a 4.7% gain (8.26/7.89) for a 10.8% clock speed gain (4.1/3.7). Or 4.35% for a 10% speed gain. Multiply by 2 for the 20% gain AMD stated, 8.7%.

So Kaveri @ 3.7Ghz would be 7.89 x 1.087 = 8.58 which is darn close to the 8.53 number by AMD. No need to add a further percentage to the number AMD provided.

As many suspected the IPC gains while great are being hampered by the clock speed reduction for traditional workloads. The GPU gains with GCN are where the big gains are to be found.

Linearity was not used. I don't know what are doing. Kaveri got 20% more score in X264 than the Richland at 3.7Ghz, therefore you only need to multiply the 7600 score in the above benchmark by 1.20:

7.89 x 1.20 = 9.5

 

[juanrga]

So basically kaveri will finally allow an APU to reach the phenom II x4 level of performance.

Where is your silly prediction that Kaveri would be a 37% slower than Bulldozer?

 

[-Fran-]

So basically kaveri will finally allow an APU to reach the phenom II x4 level of performance.

Take a look at the FX4300 and the A8 3850. I would say Phenom II was officially left in the dust with PDs introduction and new software.

Cheers!

 

[Cazalan]

Linearity was not used. I don't know what are doing. Kaveri got 20% more score in X264 than the Richland at 3.7Ghz, therefore you only need to multiply the 7600 score in the above benchmark by 1.20:

7.89 x 1.20 = 9.5

The scaling in that benchmark just wasn't making sense to me. 10% IPC should be similar to 10% clock speed. So I was trying to make sense of it. Looking further at the fine print, AMD was using 5.0.1 of the x264 benchmark not 5.0.

http://us.hardware.info/reviews/4993/15/intel-core-i3-4330--i5-4440-review-affordable-haswells-benchmarks-igpu-tech-arp-x264-hd-501-pass-2

This one makes more sense than the 5.0 results. AMD still got lower numbers somehow but:
7.6 x 1.1 = 8.36 (closer to the 6800K number)
7.6 x 1.2 = 9.12 (Kaveri @ 3.7Ghz) puts it within 11% of the 2500K

 

[etayorius]

So basically kaveri will finally allow an APU to reach the phenom II x4 level of performance.

Passmark reports Richland to be 20% Faster than the Phenom II 980 in Single Core Benchmarks, So i think Kaveri should be up to 40% faster than the 980 in single core performance, and about 20% or faster in Multi Core performance, so to answer your question... NO, Kaveri should be faster overall than the 980 in both Single and Multi Core performance.

I would not trust any benchmarks with Cinebench, that crap it´s biased in all regards... no one should use that crap to bench CPUs.

 

[juanrga]

mantle is 45% faster.

On a specific title, one a specific hardware config, on a specific OS.

On a specific universe as well.

 

[blackkstar]

So basically kaveri will finally allow an APU to reach the phenom II x4 level of performance.

This meme needs to die. "AMD finally going to catch up to their ancient architecture! xD"

If you're running code that uses the full instruction sets of both CPUs, Piledriver will obliterate K10.

Juanrga Sorry man I have no idea how the hell I think your name is juaranga. I've been doing that forever and I don't know where it comes from. I don't have problems with other words but your name just makes my brain do nothing but spin in circles and fart.

 

[juanrga]

Linearity was not used. I don't know what are doing. Kaveri got 20% more score in X264 than the Richland at 3.7Ghz, therefore you only need to multiply the 7600 score in the above benchmark by 1.20:

7.89 x 1.20 = 9.5

The scaling in that benchmark just wasn't making sense to me. 10% IPC should be similar to 10% clock speed. So I was trying to make sense of it. Looking further at the fine print, AMD was using 5.0.1 of the x264 benchmark not 5.0.

http://us.hardware.info/reviews/4993/15/intel-core-i3-4330--i5-4440-review-affordable-haswells-benchmarks-igpu-tech-arp-x264-hd-501-pass-2

This one makes more sense than the 5.0 results. AMD still got lower numbers somehow but:
7.6 x 1.1 = 8.36 (closer to the 6800K number)
7.6 x 1.2 = 9.12 (Kaveri @ 3.7Ghz) puts it within 11% of the 2500K

The hardwarecanucks benchmark puts Kaveri within the 8% of the 2500k (10.3 vs 9.5). The difference between 8% in one benchmark and 11% in another benchmark is a mere 3%, which is within the margin of error.

The point is that Kaveri CPU is at the i5-2500k level, as some here predicted

Note that I did that simulation before Kaveri was presented at APU13 and then I assumed 4GHz. Correcting for 3.7GHz, Kaveri would score something as 94.35, which is within a 8% of the 2500K.

 

[noob2222]

@juan

You never could understand that ES on cosmology is not Kaveri could you. Or that the blog calculations were 100% fake.

As for my phenom statement, look at the x264 chart provided. Richland is slower than the 980 x4 so kaveri will finally catch up. Comparing fx to phenom is no question but I have said before that even richland can be considerably slower than its PD counterpart. APU ! = FX, kaveri != FX. APU cores are weaker than CPU cores in order to make room for the GPU.

 

[esrever]

Already know from where AMD got the 20% IPC for Steamroller claimed this week-end. It is from using 2nd pass of the x264 benchmark. Kaveri scored 8.53 and Richland 7.09. Both clocked at 3.7GHz under W8.1 and 2133 memory.

Taking this review as base

Kaveri would score ~9.5, which is at the i5-2400 level and only ~8% behind FX-6300 and i5-2500k. It looks very close to my prediction ;-)

We will see up to 40% IPC gains in other benchmarks.

My prediction about HSA boost has been completely demolished. I took a 3x gain as conservative estimation

but I was expecting to see at best 5x boost in LibreOffice. How wrong I was... LibreOffice Calc run >8x faster with HSA enabled!

Some of your guesses are close, so what? Just because I guess broadwell will offer about 5% increase in performance doesn't mean I would be wrong. It doesn't show anything. As for h.264, its unlikely to be 20% faster simply due to no improvement on L2 cache and the APUs already scaling pretty well on it considering the increase in the instruction throughput. I would guess closer to slightly over 15% IPC and slightly more than 10% actual performance.

 

[logainofhades]

I don't think it is so much to make room as it is the consumer segment the APU is targeted it at. They are for average, more mainstream use. FX are enthusiast class parts. The thing that hurt APU's the most, is lack of an L3 cache.

 

[jdwii]

I don't think it is so much to make room as it is the consumer segment the APU is targeted it at. They are for average, more mainstream use. FX are enthusiast class parts. The thing that hurt APU's the most, is lack of an L3 cache.

Well since Amd may not be producing anymore FX processors for the next 2 years based on their current road maps it does matter to a lot to people that this processor does compete in that level of performance.

 

[FALC0N]

AMD has never said they wont be producing any new FX processors in the next 2 years.

 

[de5_Roy]

AMD prototype PC could be mistaken for an envelope
http://www.pcworld.com/article/2084883/amd-prototype-pc-could-be-mistaken-for-an-envelope.html
moar details on 'project discovery' ffrd, powered by

Spoiler

[strike]temash's arm replacement[/strike]

mullins soc.

amd d*ck port is now and extension of dp
http://www.vesa.org/featured-articles/displayport-adds-dockport-extension-to-royalty-free-vesa-standard/

CYBERPOWERPC Debuts Zeus Mini Small Form Factor Series
http://www.techpowerup.com/196592/cyberpowerpc-debuts-zeus-mini-small-form-factor-series.html

AMD has never said they wont be producing any new FX processors in the next 2 years.

that's true. because the moment they admit that, fx sales will drop even further from the current small volumes. kaveri launch won't make fx sales better (quite the opposite). amd's current goal is to get people buy as much fx as possible, that's part of the reason why fx cpus go on sale, prices are so low (fx customers win). the main reason is that intel keeps constant pressure on them.

 

[Master-flaw]

I think they're putting a halt on it until they get more software support. Can't blame them really, the benchmarks and PR from Intel is really putting them over...not saying it's gunna be 2 years but I don't think they even know how long...
They are doing well with introducing Free-sync and lowering the prices of their hardware...thing is, how long can they do this for...if they release a new line they have to make money off of it and with more support they will be able to price it respectively.

 

[juanrga]

Spoiler

Already know from where AMD got the 20% IPC for Steamroller claimed this week-end. It is from using 2nd pass of the x264 benchmark. Kaveri scored 8.53 and Richland 7.09. Both clocked at 3.7GHz under W8.1 and 2133 memory.

Taking this review as base

Kaveri would score ~9.5, which is at the i5-2400 level and only ~8% behind FX-6300 and i5-2500k. It looks very close to my prediction ;-)

We will see up to 40% IPC gains in other benchmarks.

My prediction about HSA boost has been completely demolished. I took a 3x gain as conservative estimation

but I was expecting to see at best 5x boost in LibreOffice. How wrong I was... LibreOffice Calc run >8x faster with HSA enabled!

Some of your guesses are close, so what?

So the trolls that said that the numbers were fake and that it was impossible for AMD to achieve that level of performance were wrong. One of those trolls even PM for months pretending me to change the numbers to fit his.

AMD has never said they wont be producing any new FX processors in the next 2 years.

AMD has never said they won't be producing a 11-core Phenom V in the next 2 years. Will they make one? No.

 

[Reynod]

So basically kaveri will finally allow an APU to reach the phenom II x4 level of performance.

noob take a hike ... there is no need to troll here.

Next time you won't get a warning ... you will be absent for a week.

 

[gamerk316]

In other news, more info on Freesync:

http://www.pcper.com/reviews/Graphics-Cards/AMD-Variable-Refresh-FreeSync-Could-Be-Alternative-NVIDIA-G-Sync

So in short:

1: Embedded Display Port 1.3, launched in 2013, has a mechanism for supporting variable refresh rates.
2: AMD is pushing for this feature in the next version of Display Port (1.3).
3: The majority of monitors, at present, do not support variable refresh rates. In theory, you could do this via firmware update, but for end users, this still almost certainly requires getting a new monitor.
4: Will only work using a native Display Port 1.3 connection.

So it comes down to the panel makers really, to see which tech gets supported first. NVIDIA has the edge, since DP 1.3 won't even be standardized until March/April, and panels based off the won't be released until the later half of the year. Meanwhile, Gsync can be integrated today. Advantage NVIDIA.

And I STILL want to know how AMD is handling setting up the Vblank intervals directly when you can't know the FPS in games ahead of time in a way that doesn't introduce lag into the system.

EDIT

NVIDIA's response:

http://techreport.com/news/25878/nvidia-responds-to-amd-free-sync-demo

*popcorn*

 

[Reynod]

http://www.fudzilla.com/home/item/33587-amd-formally-unveils-kaveri

I apologise if this has already been posted.