Vishera spotted running at 5GHz

[soccerballtux]

I disagree. 4Ghz Piledriver cores are probably marginally better than Core 2 Duo cores running at 4ghz.

 

[Mallibu]

I disagree. 4Ghz Piledriver cores are probably marginally better than Core 2 Duo cores running at 4ghz.

I highly doubt that since a FX 8150 core@ 4ghz is worse than a i7 920 core @ 2.6 ghz. Phenom also was slower core to core and at same ghz than a Core 2 quad core, and FX is worse than phenom.

 

[lifeblood]

Xbit labs is reporting an Oct 23 a launch date.

http://www.xbitlabs.com/news/cpu/display/20120921231354_AMD_Sets_the_FX_Vishera_Launch_Date.html

Finally, we have a date (assuming their correct).

 

[inf64]

I highly doubt that since a FX 8150 core@ 4ghz is worse than a i7 920 core @ 2.6 ghz. Phenom also was slower core to core and at same ghz than a Core 2 quad core, and FX is worse than phenom.

Phenom (65nm) was about 6.2% slower vs C2q(65nm). Phenom II was 2.1% slower vs C2Q(45nm). FX is worse than Phenom in certain workloads and better than Phenom in certain workloads. Nothing is black and white,especially wen Bulldozer is in question.

And 4Ghz BD/PD will destroy 2.6Ghz i7 920 in almost any scenario(yes even single threaded cases since the clock difference is more than 50% and IPC difference is around 40-45% in ST workloads). In application performance even 8150 is faster than 2.6Ghz i7 920. In MT workloads there is no contest between 4Ghz BD/PD and 2.6Ghz i7 920 . Unless you are an intel fanboi that is. In that case you may even think that pentium(c2q) is faster than 4Ghz bulldozer

 

[Makaveli]

Phenom (65nm) was about 6.2% slower vs C2q(65nm). Phenom II was 2.1% slower vs C2Q(45nm). FX is worse than Phenom in certain workloads and better than Phenom in certain workloads. Nothing is black and white,especially wen Bulldozer is in question.

And 4Ghz BD/PD will destroy 2.6Ghz i7 920 in almost any scenario(yes even single threaded cases since the clock difference is more than 50% and IPC difference is around 40-45% in ST workloads). In application performance even 8150 is faster than 2.6Ghz i7 920. In MT workloads there is no contest between 4Ghz BD/PD and 2.6Ghz i7 920 . Unless you are an intel fanboi that is. In that case you may even think that pentium(c2q) is faster than 4Ghz bulldozer

In games Bulldozer will have trouble keeping the distance.

Example look at the borderlands 2 benchmarks from Techspot a stock 920 vs a stock 8150 has the gen 1 i7 10% faster. Add another 400mhz to the amd chip isn't going to provide a big enough boast for it to lead.

Now as for other applications yes the gap may favour BD. However if you clock that 920 to 4Ghz I know which chip I would rather have!

 

[inf64]

Yes,in Borderlands 2 (wow a game!) BD has no chance to match it. Who cares exactly? Whomever wants to only play games on his PC with constant >60fps in ANY game will probably buy 3770K or 39xxK with SLI 680s and won't look back. Those people represent (roughly) just a couple of percents of client market. The rest will buy something much cheaper or something that has much better perf./$. That's where the money is.

 

[Mallibu]

Phenom (65nm) was about 6.2% slower vs C2q(65nm). Phenom II was 2.1% slower vs C2Q(45nm). FX is worse than Phenom in certain workloads and better than Phenom in certain workloads. Nothing is black and white,especially wen Bulldozer is in question.

And 4Ghz BD/PD will destroy 2.6Ghz i7 920 in almost any scenario(yes even single threaded cases since the clock difference is more than 50% and IPC difference is around 40-45% in ST workloads). In application performance even 8150 is faster than 2.6Ghz i7 920. In MT workloads there is no contest between 4Ghz BD/PD and 2.6Ghz i7 920 . Unless you are an intel fanboi that is. In that case you may even think that pentium(c2q) is faster than 4Ghz bulldozer

FX core < Phenom II core
Phenom II core < Core 2 Quad core
Core 2 Quad core < I7 core
However the discussion was on if FX is slower than Core 2 Quad core to core and clock to clock, and it is. Kinda shows how sad things are for AMD though.

Yes,in Borderlands 2 (wow a game!) BD has no chance to match it.

In a lot of games, not just one game.

Who cares exactly?

People that play games aka a lot.

Whomever wants to only play games on his PC with constant >60fps in ANY game will probably buy 3770K or 39xxK with SLI 680s and won't look back. Those people represent (roughly) just a couple of percents of client market. The rest will buy something much cheaper or something that has much better perf./$. That's where the money is.

That's why people buy i5 2500 & a single 670/7950 GPU..because you achieve >60fps with half the money spent in comparison with the rig you posted. Hyperbole is bad.

 

[Makaveli]

Yes,in Borderlands 2 (wow a game!) BD has no chance to match it. Who cares exactly? Whomever wants to only play games on his PC with constant >60fps in ANY game will probably buy 3770K or 39xxK with SLI 680s and won't look back. Those people represent (roughly) just a couple of percents of client market. The rest will buy something much cheaper or something that has much better perf./$. That's where the money is.

The point of using borderlands two is just one example I can show you more real world benchmarks of a 920 at 4Ghz beating a bulldozer overclocked in none gaming.

I shall however leave you two to finish up.

 

[inf64]

Nothing to finish up with that poster. He is just baiting others into pointless discussions. I linked the hardware.fr charts and everybody can see the application performance of FX8150. It's a bit slower than 2600K (12%). In games Bulldozer is no match for i5 or i7 but it does a solid job nonetheless. Who primarily games on his PC will buy intel for sure. For other tasks (an actual productive work,not BS games), FX is better than 2500K as can be seen in the link. Both at stock of course. Only a fragment of client PC buyers mess around with their settings (OCing,tuning) so stock vs stock AMD is positioned well where the money is (sweet spot segment at around 150-250$). FX8350 will just strengthen this position a bit(compared to FX8150) since intel upped the bar with i5s based on IB.

 

[guskline]

Fianlly a release date. I want to read the matchup with a 1100 Thuban and a 8150 first. Then on to the 2500k/2600k/3570k and 3770k.

My predictions?

Obviously at stock, the 8350 will beat the 8150 ( Heck Homer Simpson could even figure that out!)

I would OC the 8150 to 4Ghz and then compare stock 8350 to it Match the 8150, 1100T and the 8350 at 4 Ghz and on up. Then match the 8350 stock to the 2500k/2600k/3570k and 3770k.

The 8350 will edge out the 8150 due to better IPC and "might" OC higher than the 8150 due to Resonant Mesh tech but I think it will struggle to beat the 2600k. Doubt it will touch the 3770k.

Final prediction AMD will sound ther trumpets with the first 4 GHZ stock multi core cpu.

 

[Mallibu]

Nothing to finish up with that poster. He is just baiting others into pointless discussions. I linked the hardware.fr charts and everybody can see the application performance of FX8150. It's a bit slower than 2600K (12%). In games Bulldozer is no match for i5 or i7 but it does a solid job nonetheless. Who primarily games on his PC will buy intel for sure. For other tasks (an actual productive work,not BS games), FX is better than 2500K as can be seen in the link. Both at stock of course. Only a fragment of client PC buyers mess around with their settings (OCing,tuning) so stock vs stock AMD is positioned well where the money is (sweet spot segment at around 150-250$). FX8350 will just strengthen this position a bit(compared to FX8150) since intel upped the bar with i5s based on IB.

The two summarizing graphs from the site you posted: (http://www.hardware.fr/articles/863-22/moyennes.html)

FX at 150.7
i5 3570 at 148.8
Hardly a difference in the productive apps.

And the 2nd:

FX: 106
i5 3570: 172
Now, that's a difference.
But I like the fact that you pointed out the 1% difference out of the first graph, and completely ignored 50% difference in second. Cherry picking at it's finest.

 

[inf64]

I said I won't reply but here I must address your point since you apparently cannot read. I distinctively mentioned application performance and gaming performance. And I said intel is better in games but you failed at reading I guess. 3570K in actual application workloads is performing worse and costs 40$ more (newegg prices as of now).

 

[Racan]

And I said intel is better in games but you failed at reading I guess. 3570K in actual application workloads is performing worse and costs 40$ more (newegg prices as of now).

For application workloads I think Intel offers a good alternative to the 3750k(same price) and Bulldozer with the Ivy Bridge based 4c/8t E3-1230V2 Xeon.

http://www.pcgameshardware.de/CPU-Hardware-154106/Tests/Xeon-E3-1230-v2-im-Test-907778/

You can't overclock it though.

 

[MLSCrow]

I see that it was already posted, but yeah, seems like October 23rd is the release date. On top of that, I didn't see this picture before, but you can see in the background that the FX8350 can score above 9.0pts in Cinebench, which is roughly what a 2600K would score at ~4.6GHz. Not great, but not bad either. There's no way to actually see what speed it was set to in order to achieve that 9.06, but I'm assuming it was either the 4.82 that they are showing in the CPUz frame or it was the 5.0GHz they reported boosting it to at the show. Usually Cinebench highlights your current score, but I don't see anything highlighted, so who knows? Either way, not too bad.

 

[AtenRa]

I see that it was already posted, but yeah, seems like October 23rd is the release date. On top of that, I didn't see this picture before, but you can see in the background that the FX8350 can score above 9.0pts in Cinebench, which is roughly what a 2600K would score at ~4.6GHz. Not great, but not bad either. There's no way to actually see what speed it was set to in order to achieve that 9.06, but I'm assuming it was either the 4.82 that they are showing in the CPUz frame or it was the 5.0GHz they reported boosting it to at the show. Usually Cinebench highlights your current score, but I don't see anything highlighted, so who knows? Either way, not too bad.

http://forums.anandtech.com/showpost.php?p=33992066&postcount=58

 

[serpretetsky]

that is so not fair...

4ghz 5 yrs ago is not 4ghz today.

Infact any modern processor now at 4ghz would spin circles arround any of my conroe's which did 4ghz without even breaking a sweat and asking it why its struggling.

That is slightly out of context. His post was a direct reply to somehow who said AMD was the first company to break 4ghz on production CPU's. He was pointing out that other companies have already done this. He wasn't talking about overall performance at all. I can't say whether or not he's right, I'm not familiar with 4ghz production cpu's.

 

[AtenRa]

http://www.youtube.com/watch?feature=player_embedded&v=o5hxOsrukPY

+7% IPC

+7% More Frequency

+15% More efficient

 

[Smoblikat]

http://www.youtube.com/watch?feature=player_embedded&v=o5hxOsrukPY

+7% IPC

+7% More Frequency

+15% More efficient

Pathetic. The arch is 7% faster.

 

[inf64]

Pathetic. The arch is 7% faster.

The "arch" is basically the same Bulldozer on the same process node. Do you know how hard is to even get 1% of "IPC" improvement in integer workloads? It takes literary dozen of of changes to various parts of the core.
7% is perfectly fine for a 1 year "quick fix". It's a tweaked old BD core and now it became more efficient. For more we will have to wait 1 more year.

 

[Makaveli]

7% is decent its what you got from SB to IVY.

If you were to compare it to intels releases this is basically just a tick.

Anyone expecting this to just equal SB or IVY in that short of time is just loco.

It will take amd many years to close the IPC gap.

For those with AMD's systems it may be a worthy update.

 

[sequoia464]

7% is decent its what you got from SB to IVY.

For those with AMD's systems it may be a worthy update.

With the 15% efficiency increase (hoping this means lower max watts) I might consider getting one - I have a 990FX board already.

Another couple of weeks and we will know for sure - I'm speculating we will also get a few new epic threads about how good/bad piledriver actually is.

 

[pelov]

The "arch" is basically the same Bulldozer on the same process node. Do you know how hard is to even get 1% of "IPC" improvement in integer workloads? It takes literary dozen of of changes to various parts of the core.
7% is perfectly fine for a 1 year "quick fix". It's a tweaked old BD core and now it became more efficient. For more we will have to wait 1 more year.

It makes me wonder just how much there was to "quick fix" in the original Bulldozer core. It means that the original Bulldozer module approach must have been pretty horrible. Considering there's already a laundry list of improvements in Piledriver, the Steamroller improvements should probably net a decent gain.

Does anyone know which fabs and what process/node AMD has planned for excavator? I believe Steamroller is supposed to be 28nm GloFo bulk

 

[AtenRa]

Does anyone know which fabs and what process/node AMD has planned for excavator? I believe Steamroller is supposed to be 28nm GloFo bulk

If im not mistaken same 28nm GloFo for Excavator.

 

[guskline]

Question: Is the 7% improvement measured from a 3.6Ghz Bulldozer to a 4 Ghz PildeDriver or did they run the CPUs at the same clockspeed?

 

[guskline]

I doubt it's worth it if you already have a 8150.