Anyone else eyeing AMD's Carizzo?

AMD is on the right track. Improving efficiency is good, still have some ways to go though.
 
If there's any truth to the onchip graphics being twice as fast as the R5 based graphics, I'd be all over this so long as they retain the same socket. I just upgraded to an FM2+ mobo in my HTPC, I'm not upgrading again.
 
Same, I have a really nice fm2 board and this cpu at 4.6+ I WANT! lol I can't wait to see what this can do.
 
I'm confused. The comparison was to a 3.2 GHz Kaveri CPU (Steamroller) was it not? Achieving the same performance level while being 600 MHz slower shows IPC improvements, which are a good thing for AMD, is it not?
 
Completely out of touch with AMD and what they have going on for CPU's but VERY happy to see this. I wasn't eyeing this before but I will be now. Thanks OP.
 
not too interested.

it's a mobile product, they haven't even outfitted it with a full compliment of PCIe lanes for the GPU.

it will be a fine mobile product tho... just nothing i'm interested in.
 
Finally. Some decent gains in CPU performance. Hopefully this source of info is right. I also hope AMD gets these cores up to the high end line as soon as they finish testing the waters with them in the APU's.
 
yep, excavator marks the end of the bulldozer saga and the beginning of Jim Keller's awesome cpu designs. I'm stoked
 
Carrizo 15W;
http://www.sisoftware.eu/rank2011d/...d4e2d4e3d7e4d2f486bb8badc8ad90a086f5c8f0&l=en

Kaveri 20W
http://www.sisoftware.eu/rank2011d/...d4e1d6e1d1e8dff98bb686a0c5a09dad8bf8c5fd&l=en



Well I did a bit of number crunching on those sisoft scores.

The Kaveri cpu at 3ghz 4T gets: 24.2GOPS
The Carrizo cou at 2.6ghz 4T gets: 24.7GOPS

Kaveri- 0.0080666666666667 GOPS per mhz
Carrizo-0.0095 GOPS per mhz

Improvement 15% IPC

Compared to Haswell

Haswell 4690k CPU@ 3.5ghz 4T gets: 92.6 GOPS
Haswell- 0.0264571428571429 GOPS per mhz


Speculation numbers: @ 4 threads/cores
Carrizo@ 4ghz- 38 GOPS
Kaveri@4ghz- 32.27 GOPS
Haswell@4ghz- 105.83 GOPS

Haswell@4ghz @ 2 threads/cores ( I just divided the 4T/4C results by 2, as I was to lazy to find a dual core haswell score)
52.915 GOPS

OR Haswell is 2.79 times faster at the same clock speed given all the CPU's are 4 Core/ 4 Thread. This means that 4 core Intel's even with out hyper threading are still way outta reach for AMD. However the Dual core parts are a bit threatened depending on the clock speeds. Granted this is just one benchmark, but it doesn't really move AMD position at all. Granted the parts here are 15-20watt AMD parts compared to the 84 watt haswell. On the mobile side of things seem Carrizo shows some nice improvements. We will see what kinda SKU's Amd releases and what final clock speeds and such will be but they have a chance to improve there ULV designs by good margin.
 
ebduncan said:
Carrizo 15W;
http://www.sisoftware.eu/rank2011d/...d4e2d4e3d7e4d2f486bb8badc8ad90a086f5c8f0&l=en

Kaveri 20W
http://www.sisoftware.eu/rank2011d/...d4e1d6e1d1e8dff98bb686a0c5a09dad8bf8c5fd&l=en



Well I did a bit of number crunching on those sisoft scores.

The Kaveri cpu at 3ghz 4T gets: 24.2GOPS
The Carrizo cou at 2.6ghz 4T gets: 24.7GOPS

Kaveri- 0.0080666666666667 GOPS per mhz
Carrizo-0.0095 GOPS per mhz

Improvement 15% IPC

Compared to Haswell

Haswell 4690k CPU@ 3.5ghz 4T gets: 92.6 GOPS
Haswell- 0.0264571428571429 GOPS per mhz


Speculation numbers: @ 4 threads/cores
Carrizo@ 4ghz- 38 GOPS
Kaveri@4ghz- 32.27 GOPS
Haswell@4ghz- 105.83 GOPS

Haswell@4ghz @ 2 threads/cores ( I just divided the 4T/4C results by 2, as I was to lazy to find a dual core haswell score)
52.915 GOPS

OR Haswell is 2.79 times faster at the same clock speed given all the CPU's are 4 Core/ 4 Thread. This means that 4 core Intel's even with out hyper threading are still way outta reach for AMD. However the Dual core parts are a bit threatened depending on the clock speeds. Granted this is just one benchmark, but it doesn't really move AMD position at all. Granted the parts here are 15-20watt AMD parts compared to the 84 watt haswell. On the mobile side of things seem Carrizo shows some nice improvements. We will see what kinda SKU's Amd releases and what final clock speeds and such will be but they have a chance to improve there ULV designs by good margin.
Click to expand...
Do you mind running those numbers again with a now 4 year old Sandy Bridge part?
I seem to recall the old comparison was a 4C, 4T i5-2500k was roughly twice as fast per core as bulldozer at the same clock speed, and the gap only narrowed due to the extremely high speeds AMD chips were running at.

Problem here is the fact that Intel has improved their chip's IPC by approximately 15% also, so AMD's improvement still means little. What they need now is another 50% jump. At least they are heading in the right direction. I would like to see AMD at least reach Sandy Bridge level of performance (IPC wise, per core, per clock) as the current concensus is that CPU bottlenecks are occuring under certain situations with Intel CPU's in generations prior to Sandy Bridge and Westmere.
 
CaptNumbNutz said:
Do you mind running those numbers again with a now 4 year old Sandy Bridge part?
I seem to recall the old comparison was a 4C, 4T i5-2500k was roughly twice as fast per core as bulldozer at the same clock speed, and the gap only narrowed due to the extremely high speeds AMD chips were running at.

Problem here is the fact that Intel has improved their chip's IPC by approximately 15% also, so AMD's improvement still means little. What they need now is another 50% jump. At least they are heading in the right direction. I would like to see AMD at least reach Sandy Bridge level of performance (IPC wise, per core, per clock) as the current concensus is that CPU bottlenecks are occuring under certain situations with Intel CPU's in generations prior to Sandy Bridge and Westmere.
Click to expand...

Sandybridge in the same benchmark is

0.0154047619047619 GOPS per Mhz

Ivy Bridge is the same benchmark is

0.0165454545454545 GOPS per Mhz

So at 4ghz like the others above 4 cores/ 4 thread
Sandy- 61.619 GOPS
Ivy- 66.181 GOPS

Keep in mind this is Sandra results do vary quite a bit, and it doesn't represent any real world applications.
 
^Not in ST on average. About 5% faster on average for ST compared to Ivy. Can get bigger boosts than that if using a program that leverages AVX2 and such. Some programs saw pretty big boosts, like the Dolphin emulator.
 
NaroonGTX said:
^Not in ST on average. About 5% faster on average for ST compared to Ivy. Can get bigger boosts than that if using a program that leverages AVX2 and such. Some programs saw pretty big boosts, like the Dolphin emulator.
Click to expand...

Would that explain the %40 gain in the benches mentioned earlier?
 
Just to be clear its just one benchmark.

40% does seem like a-lot, but Haswell made some good improvements to the CPU uarch and it shows in this bench.

On average Haswell is 15% faster per clock than Ivy-bridge across multiple benchmarks/apps

I wouldn't go by Sandra results as the do all end all. Give it some time and more information will come available on AMD's Carrizo cpus and how they stack against Intel offerings in multiple benchmarks.
 
ebduncan said:
Just to be clear its just one benchmark.

40% does seem like a-lot, but Haswell made some good improvements to the CPU uarch and it shows in this bench.

On average Haswell is 15% faster per clock than Ivy-bridge across multiple benchmarks/apps

I wouldn't go by Sandra results as the do all end all. Give it some time and more information will come available on AMD's Carrizo cpus and how they stack against Intel offerings in multiple benchmarks.
Click to expand...

Just so it's clear Haswell is also less than 1% faster than ivy bridge in some benchmarks. I can post some truecrypt benchmarks if people are interested.
 
given that recent intel architectures (and very recent AMD ones too), do their AES encoding/decoding on the custom AES/NI logic I really wouldn't expect there to be any difference in truecrypt performance.
 
R3MF said:
given that recent intel architectures (and very recent AMD ones too), do their AES encoding/decoding on the custom AES/NI logic I really wouldn't expect there to be any difference in truecrypt performance.
Click to expand...

Not always the case.

http://www.legitreviews.com/wp-content/uploads/2014/06/truecrypt2.jpg
truecrypt2.jpg
 
Anyone else inevitably read it as Chorizo every time?

If it were chorizo, I'd build a system just to be able to say my computer ran on chorizo.
 
Modred189 said:
Anyone else inevitably read it as Chorizo every time?

If it were chorizo, I'd build a system just to be able to say my computer ran on chorizo.
Click to expand...

Not really. I do not know what Chorizo even is but I am looking forward to their final release of the CMT architecture.
 
So, at the same clock, it is 23% faster than Kaveri.

Still would have some serious catching up to do to match Intel's performance, unless it can clock much higher too, but that is not very likely since they have shifted from SOI to SHP process which doesn't lend itself to clocking as high on CPU's (unless they plan on undoing this...)
 
Carrizo won't be on the same process as Kaveri was (28nm SHP BULK). It will still be 28nm, but I don't know exactly on which node. Rumors say "GF28A" but that is a family of nodes last I checked.
 
schmuckley said:
It's Fudzilla :rolleyes:
not [H]
I'd trust Coolaler more.
Click to expand...

It is hardware.info which has the source article , but it is speculation. Never the less it is something they found it useful commenting on.

A news article with comments in it is always speculation and the "original" Dutch version differs from the stuff that was posted by Fudzilla.

Factual information trumps 3rd rate commentary upon some benchmarks. If that was only the case for most tech websites then we would only have just a few reliable sites left.
 
Modred189 said:
Anyone else inevitably read it as Chorizo every time?

If it were chorizo, I'd build a system just to be able to say my computer ran on chorizo.
Click to expand...

YES! I was just thinking the same damn thing.......




ebduncan said:
No



Chorizo is sausage.
Click to expand...

Sense of humor much?

We will see, I'm excited for it as I mostly rock laptops. If it is anywhere near intel's mobile CPU's when they are released I'll be buying a laptop with one.
 
So no mention of the desktop APU availbiltiy...3rd qtr maybe? That is a long time to wait for an updated less slower APU...does not sound good.
 
You must log in or register to reply here.
Back
Top