So.....what are you really expecting from Zambezi cpus?

[386DX]

IPC will be much slower then Sandy Bridge and about 5-10% faster then Phenom II. However a 6 Core BD will be slower then a 6 Core Phenom II at the same clock speed due to penalty for shared resources per module.

AMD will use turbo and more cores to make the benchmarks competitive. This is my prediction on performance:

Single-Thread:
2600K > 2500K > FX-8150 > FX-8100/8120 > FX-6100 > FX-4100 > Phenom II

Multi-thread:
FX-8150 > 2600K > FX-8100/8120 > 2500K >= Phenom II > FX-6100 > FX-4100

At typical 24/7 air overclock speed the 2600K and 2500K will be faster then BD at everything.

 

[inf64]

inf64, if that was true, wouldn't Cray refuse shipment?

Edit: Wow, Vesku said almost the exact same thing in another thread... Great minds...

At least I know I'm not the only person who finds it hard to believe BD is such a hard fail that AMD is slipping backwards (and that their partners don't care while AMD outright lies about it).

There is still this "discrepancy" between quoted server (HPC) performance increase over MC and those poor SIMD/FP scores that apparently Zambezi shows. Server statements point to ~10% better per core and per clock or at the worst case at least even per core and per clock-worst case is that AVX speedup of around 10% is figured in the 35% number(2.3Ghz 16C Interl. vs 12C 2.5Ghz MC).
This(same FP IPC per core) would imply that Zambezi @ 3.3Ghz (no Turbo!) should at least be 33% faster than 1100T Thuban or 1.33x3.6/3.3=1.45 or 45% faster in case of FX8150 vs 1100T. Unfortunately we don't see this in any of the leaks. We see the opposite . FX8150 can't match or barely matches Thuban while having 33% more cores( using AMD's own terminology of core here). Sometimes it even fails against Deneb X4.

 

[Idontcare]

Mega-tasking!?

Haven't heard that term since...

http://www.amd.com/us/press-releases/Pages/Press_Release_114483.aspx

That's the previous Quad FX (aka "QuadFather") marketing "catch-phrase".

Oy vey!

 

[exar333]

IPC will be much slower then Sandy Bridge and about 5-10% faster then Phenom II. However a 6 Core BD will be slower then a 6 Core Phenom II at the same clock speed due to penalty for shared resources per module.

AMD will use turbo and more cores to make the benchmarks competitive. This is my prediction on performance:

Single-Thread:
2600K > 2500K > FX-8150 > FX-8100/8120 > FX-6100 > FX-4100 > Phenom II

Multi-thread:
FX-8150 > 2600K > FX-8100/8120 > 2500K >= Phenom II > FX-6100 > FX-4100

At typical 24/7 air overclock speed the 2600K and 2500K will be faster then BD at everything.

2500k already spanks PhII, this is dubious, at best.

 

[LOL_Wut_Axel]

IPC will be much slower then Sandy Bridge and about 5-10% faster then Phenom II. However a 6 Core BD will be slower then a 6 Core Phenom II at the same clock speed due to penalty for shared resources per module.

AMD will use turbo and more cores to make the benchmarks competitive. This is my prediction on performance:

Single-Thread:
2600K > 2500K > FX-8150 > FX-8100/8120 > FX-6100 > FX-4100 > Phenom II

Multi-thread:
FX-8150 > 2600K > FX-8100/8120 > 2500K >= Phenom II > FX-6100 > FX-4100

At typical 24/7 air overclock speed the 2600K and 2500K will be faster then BD at everything.

I'd say this given what we know:

Single-threaded:
2600K>2500K>2400>FX-8150>FX-8120=FX-6100=FX-4100>X6 1100T

Multi-threaded:
2600K=FX-8150>FX-8120>2500K=X6 1100T=FX-8100>2400>FX-6100>FX-4100

 

[mosox]

I'm only interested in those CPUs not having hardware DRM like Intel. I like to have control over what I'm doing on my PC.

If the Intel HW DRM kicks in tomorrow the 2500K will be selling for like $20 second hand, lots of dudes have stuff on their HDDs that's worth many times more than any CPU ever built.

 

[LOL_Wut_Axel]

I'm only interested in those CPUs not having hardware DRM like Intel. I like to have control over what I'm doing on my PC.

If the Intel HW DRM kicks in tomorrow the 2500K will be selling for like $20 second hand, lots of dudes have stuff on their HDDs that's worth many times more than any CPU ever built.

Wat.

The 'DRM' you're speaking of is not DRM, and it's only available for CPUs with vPro and business/enterprise users. It's used so that enterprises can remotely access their systems and even completely disable them in case they're stolen.

This doesn't affect any of us.

 

[bryanW1995]

At least this poll was super dooper confusing!

Should have had an "other" category. My vote would have been: Epic Fail.

 

[PreferLinux]

I'm only interested in those CPUs not having hardware DRM like Intel. I like to have control over what I'm doing on my PC.

If the Intel HW DRM kicks in tomorrow the 2500K will be selling for like $20 second hand, lots of dudes have stuff on their HDDs that's worth many times more than any CPU ever built.

If you are talking about the disabling the chip, then as said, you don't have it. If you are meaning the other thing, then it is really no different to HDCP, which is in every modern graphics card, monitor and TV. You don't have to worry about your content.

 

[BlueBlazer]

Oy vey!

AMD may use "World Fastest CPU" as the new marketing catch-phrase.

 

[Dresdenboy]

Why AMD!? Why!?

First you bad-mouthed Intel's "glued together chips" only to then do MCM yourselves. Now you've decided to revisit the P4 D:

WTF is next? Rambus ram?

Reason for the first might be that Intel's first MCMs had 2 dies communicating via FSB instead of a more efficient connection.

Further P4 is a completely different beast than BD. FO4 delay went down much more (compared to PIII) than in case of BD vs. 10h. This at least seems to be inline with "knee of the curve" and "sweet spot" decisions (even IBM found 17-19FO4 in this case) and not the result of a hyperpipelining paper proposing ever increasing performance with longer pipelines. P4 also had 8FO4 circuits using very fast dynamic logic style (e.g. domino). BD is more a static design similar to what has begun with Nehalem at Intel (look at ISSCC papers).

And indeed, AMD will go Rambus - at least for their next GPUs supporting XDR2 as it seems

 

[BlueBlazer]

Reason for the first might be that Intel's first MCMs had 2 dies communicating via FSB instead of a more efficient connection.

Despite the FSB's "inefficient" connection, MCM'ed Core 2 Quad generation still proved to be a handful even for Quad FX (with efficient HT link), Phenom ("native quad core") and up to Phenom II X4 (clock-to-clock wise). Often the higher performance of the cores themselves (especially in IPC and architectural improvements) can be a big factor.

And indeed, AMD will go Rambus - at least for their next GPUs supporting XDR2 as it seems

And do we see return of the Quadfather? Considering dual socket enthusiast boards (such as EVGA's Classified SR2) for Intel Xeons exists and performed well (was used in breaking world record benchmarks).

 

[386DX]

2500k already spanks PhII, this is dubious, at best.

I'm comparing the highest end PhII part out.. the x6 1100T. The 2500K beats it in some multi-threaded apps and loses in others. Neither but any great margin so I consider there heavily multi-threaded performance about even.

 

[jhu]

I expect it to end world hunger and bring world peace.

 

[sequoia464]

Well...I was expecting to have one by now - along with some legitimate benchmarks to back up the purchase...

 

[BlueBlazer]

Well...I was expecting to have one by now - along with some legitimate benchmarks to back up the purchase...

At the time of this posting, there's still no announcements/news of shipping (not talking about Interlagos) and no release dates from AMD themselves (with only Q4 prediction of early to mid October to go on). One must wonder.... :hmm:

 

[Ancalagon44]

It will be interesting to see if anything is revealed on September the 19th, since that was/is supposed to be A reveal date, if not THE release date.

 

[GaiaHunter]

After last OC WR this poll needs a "Over 9000" option.

 

[BlueBlazer]

It will be interesting to see if anything is revealed on September the 19th, since that was/is supposed to be A reveal date, if not THE release date.

As with all products.... shipping comes first, before launch/release. :hmm: