Haswell vs Kaveri

The bias the tech sites are showing over this is staggering. I don't think I've seen a single site bar xbit mention the Broadwell delay yet they are all falling over themselves to go on and on about Kaveri.
 
I don't think I've seen a single site bar xbit mention the Broadwell delay yet they are all falling over themselves to go on and on about Kaveri.
I thought Kaveri was supposed to come out well in advance of Broadwell, no? I think they're just picking it up because Kaveri is/was supposed to be coming out really soon, whereas Broadwell is still a little ways off.

Anyways, you have a link on the Broadwell delay? Is it just the vrzone thing?
 
Yep from Intel roadmaps -

1095245-1371765699396428-Ashraf-Eassa.png

1095245-1371766012793147-Ashraf-Eassa.png


image.php

image.php


What bugs me about this is we're talking a week or two for AMD (AMD still maintains Kaveri ships this year) while Intels is a slip of a half, which could be a lot more. Sometimes I wish AMD would just stop giving out their roadmaps altogether, fuck the press.
 
To be fair, it's not as if Intel badly needed Broadwell to become competitive, which is probably why people don't care that much about its delay.
 
I think that the press is really milking this for a lot more than what it's worth. We had one comment from Lisa Su in January this year saying that Kaveri would launch at the end of this year. Since then we've had various AMD sources saying it would ship by the end of this year.

By this stage of the year there is no point in launching a chip before CES anyway, and the industry guys should know that. Any actual delay is counted in weeks at the most, yet it's being made out to be something much worse? Do you remember this fuss over the 2 month Ivy Bridge delay, or Haswell?
 
What I wonder is.. can Kaveri beat a Phenom II X6 1100T? :p

Dunno. I bet it beats Phenom II X4 and FX 4300, hopefully. Hoping it fares well against i3 3220, i3 Haswell and maybe i5 2300.
Badly needed is a triple module/six core Kaveri variant if they want to claim semi-high end status and consumer multimedia content creation.

Steamroller IPC is very close to Thuban and with the new instruction sets, and +1ghz clock speed, and 1 more core, and faster DDR3, it should pull quite ahead of Phenom II X6.
 
Steamroller IPC is very close to Thuban and with the new instruction sets, and +1ghz clock speed, and 1 more core, and faster DDR3, it should pull quite ahead of Phenom II X6.

One more core?

The X6 had 6 cores, as the name indicates. Kaveri only has 4.
 
Steamroller IPC is very close to Thuban and with the new instruction sets, and +1ghz clock speed, and 1 more core, and faster DDR3, it should pull quite ahead of Phenom II X6.
Are you saying each core has the same IPC at equal frequencies?
 
I think that the press is really milking this for a lot more than what it's worth.
You'll be very happy to know that your favorite unbiased websites, such as Anandtech, Hardocp, and Guru3D don't even have it on their front page right now.

Do you remember this fuss over the 2 month Ivy Bridge delay, or Haswell?
I remember as little fuss as there is today about Kaveri. Though I agree with others that it is a much bigger deal for AMD than for Intel. The latter can delay their full roadmap by a year and they'd still be ahead.
 
Are you saying each core has the same IPC at equal frequencies?

IPC is a relatively vague term (and apparently an Internet rage). There are things that even old BD does better than any K8L derivative, however those are not dominant in many (any?) relevant scenarios, and the cache hierarchy on BD is quite dingy thus making things even worse. I would still expect K8Ls to be better than SR in certain scenarios, clock-for-clock, albeit by perhaps a lower margin. Also, once again, what has us convinced that there's this out of band boost of single-threaded perf with SR? Expecting it to blow things out of the park might be unrealistic. That being said, the rumblings about quad-channel are interesting, it would make for a pretty neat dev-box if that is true.
 
Planet3Dnow uncovered some interesting details about upcoming desktop Kaveri chip!
http://www.planet3dnow.de/cms/?p=3279&preview=true

They show few scores from SiSoftSandra for 13CU(!!) version and to add confusion there are screens showing benchmark detecting 64bit DDR3 memory as well as 256bit DDR3 memory ...
I know it might be just a bug on SiSoft or BIOS part but still :smile:

Possible scores for Kaveri SPECTRE DESKTOP 100W (1305) Radeon M5 R200 832SP 600MHz

GP Processing:


  • AMD Kaveri APU R5 M200 @600 MHz (DDR3-1600) - 709.6 MPix/s
  • AMD Richland APU HD 8670D @844 MHz (DDR3-2133) - 239.3 MPix/s
Memory Bandwidth:


  • AMD Kaveri APU R5 M200 @600 MHz (DDR3-1600) - 15.0 GB/s
  • AMD Richland APU HD 8670D @844 MHz (DDR3-2133) – 12.4 GB/s
GP Cryptography:


  • AMD Kaveri APU R5 M200 @600 MHz (DDR3-1600) - 9.4 GB/s
  • AMD Richland APU HD 8670D @844 MHz (DDR3-2133) – 3.8 GB/s

 
Benchmark details say 1 device / 2 threads. So its probably some kind of Crossfire. An added Hainan (5CU) could explain the strange 13 CUs for this setup. Other Crossfire setups are recognized as 2 devices, but Hainan without display outputs could be recognized different.

Hainan with HSA capabilities could also explain the sense of another GPU below the 77mm² Oland/Mars.
 
Back
Top