Sign in with
Sign up | Sign in
Your question

AMD 45nm Deneb Sneak Peak @ HardSpell

Last response: in CPUs
Share
a c 96 à CPUs
a b K Overclocking
a b À AMD
July 11, 2008 5:03:56 PM

I wonder what mobo/chipset/south bridge they used?

Until sb750 rolls it's tough to figure where AMD will settle. The timing / PLL is killing OCs.
a c 122 à CPUs
a b À AMD
July 11, 2008 5:13:31 PM

Hmm... How doess that SuperPi data stack up vs a Q6600/Q9450 @ those clock speeds?
Related resources
July 11, 2008 5:21:17 PM

Oh come on... the Super PI is obviously a FSB (edit :lol: ) bottle neck on the phenom.. :oops: . o O (where's the nearest exit?)
July 11, 2008 5:23:43 PM

Someone else said their 4300 @ 3.2 does it in 20 seconds. So, on par with older C2Ds then? And an ES to boot
a c 122 à CPUs
a b À AMD
July 11, 2008 5:26:38 PM

From the CPU-Z shot it looks exactally the same as a Phenom 9500 but it says 45nm in the Technology.

I guess I can't wait till they have some real world benchmarks to see if there is possibly any advantage to these CPUs over the 65nm Phenoms.

Lets just hope that SOI @ 45nm is as reliable as they say it is.
July 11, 2008 5:27:00 PM

My Q6600 at 2.880 does the 1MB in 19 secs.
July 11, 2008 5:35:19 PM

Supposedly 12.5% faster than Phenom. Hope its true, and a few more tweaks, being its a ES
July 11, 2008 5:44:31 PM

JAYDEEJOHN said:
Someone else said their 4300 @ 3.2 does it in 20 seconds. So, on par with older C2Ds then? And an ES to boot


An E6300 @ 2.8Ghz does 20 seconds.

But Super Pi is grossly biased towards Core 2s, so it might be better to compare with AMD.

For instance, X2 6000+ running at 3.3Ghz gets about 25 seconds.
July 11, 2008 5:52:59 PM

Re-ran my SuperPI 1MB.

Q6600 - 2.880:



E4400 - 2.8:



My E4400 has SpeedStep disabled since that is the only way I can run it in my Linux OS. :lol: 

So ya, I agree with the E6300 doing the same.
July 11, 2008 6:11:34 PM

Maybe they used a abucus fsb? heheh
a b à CPUs
July 11, 2008 6:17:45 PM

Well, the E6750 I'm on right now (1748FSB, 437*8 for 3495MHz @ 1.4V) does it in 14.9
July 12, 2008 7:41:03 AM

I just ran a 14.437 on mine..see sig :D 
a b à CPUs
July 12, 2008 1:33:57 PM

Link to another article about the 45nm Phenoms.

Link to a translated page in a forum where the SuperPI results and screen shots were posted.

The mobo used in this test is running the 790 chipset with the SB600 southbridge. Interested in seeing how high an overclock you could get with the new SB750 southbridge.

It seems that AMD is moving in the right direction and the die shrink to 45nm is addressing the low clock speeds and the L3 cache latency. They bumped up the cache sizes to 512KB L2 and 6MB L3. It would be reasonable to see 3GHz and up on retail 45nm Phenoms. It also seems that AMD is postioning themselves to have a price/performance competitor with Nehalem.

Interesting time ahead with lots of new cpus and toys to play with.


a c 96 à CPUs
a b K Overclocking
July 12, 2008 2:38:00 PM

yomamafor1 said:
http://en.hardspell.com/doc/showcont.asp?news_id=3721

Wow.... 1.568V for 3.4Ghz. What a nice overclock.


...and they did it with an apparently retail board with a BIOS with a compiled-on date of January 2008. It seems a bit odd to me that an unreleased CPU is supported by a six-month-old BIOS on a retail board as something as much as a die shrink almost always needs an updated BIOS (look at the early AM2 boards that cannot run 65 nm K8s, for example.)

This looks too legit to call shenanigans as CPU-Z 1.46 would recognize the processor and it validated (not to mention he published pictures of the IHS) but I am a little...curious is the correct word to use here.
a b à CPUs
July 12, 2008 3:32:47 PM

HT Link has been slowed to 1800 ???

Seems they slowed it a tad but the cache latency looks to have improved.

Does that sound right MU ??
July 12, 2008 3:49:35 PM

Reynod said:
HT Link has been slowed to 1800 ???

Seems they slowed it a tad but the cache latency looks to have improved.

Does that sound right MU ??

1800 is correct with the current line up where only the 9750 and up have a HT Speed of 2000 Mhz.
July 12, 2008 4:58:46 PM

MU_Engineer said:
...and they did it with an apparently retail board with a BIOS with a compiled-on date of January 2008. It seems a bit odd to me that an unreleased CPU is supported by a six-month-old BIOS on a retail board as something as much as a die shrink almost always needs an updated BIOS (look at the early AM2 boards that cannot run 65 nm K8s, for example.)

This looks too legit to call shenanigans as CPU-Z 1.46 would recognize the processor and it validated (not to mention he published pictures of the IHS) but I am a little...curious is the correct word to use here.


I agree. Although personally I don't hold much hope for the SB750, but we will indeed wait and see.
July 12, 2008 5:40:41 PM

Hopefully AMD has a winner here. We need some competition
a b à CPUs
July 13, 2008 10:26:18 AM

Core voltage looks too high as well.

a b à CPUs
July 13, 2008 10:31:37 AM

It could be the early ES playing up tho... but remeber the original ES Phenom?
a b à CPUs
July 13, 2008 11:06:02 AM

Yes ... Though look at the core voltage for the new Penryn class cpu ... 45nm chips should run quite a bit lower.

For it to run so high suggests that the process is not so lean on current draw as AMD would have liked.

Obviously all conjecture though ... the shots could easily be fake.

a b à CPUs
July 13, 2008 12:15:22 PM

Photoshopping screenshots is easier because of the lack of shadowing and etc.
a c 122 à CPUs
a b À AMD
July 13, 2008 4:39:32 PM

Or it could be that SOI @ 45nm needs more juice for stability. We will have to wait and see once the hardware site get their hands on some ES chips for full blown testing and OCing to see what its capable of.
a b à CPUs
July 13, 2008 10:07:43 PM

They are ES chips... they are not retail... I'd wait for those as ES are normally cherry-picked...
a c 122 à CPUs
a b À AMD
July 13, 2008 11:51:37 PM

^this much is true. The fact that it OCs so well almost makes me think so as last time AMD had such high speed CPUs they were so cherry picked that not even a cherry could tell the difference between it and Phenom.

I just hope AMD doesn't do the whol "In House testing only" crap they pulled with the Phenoms original launch. That right there made me loose respect for them.
a b à CPUs
July 14, 2008 1:26:46 AM

If you think about it is quite pointless to do so. You lose respect... all it takes is launch day and billions of reviews flood in, negative. All the excitement and etc. turns into anger and etc., all before the actual purchase...

Anyways, what happened to those cherry picked Phenoms anyway? I suppose at least one in 1000 are good? why not sell those as 'special' Phenoms... use them to clock to 3Ghz?
July 14, 2008 5:41:47 AM

I believe they're currently known as "black editions", where they are the cream of the crops.
a b à CPUs
July 14, 2008 7:05:05 AM

mmm... but remeber those that beat Core 2 Quads clock for clock?
July 14, 2008 7:55:01 AM

There are Phenoms that beat C2Q clock for clock?
a b à CPUs
July 14, 2008 9:29:13 AM

those ES ones, yes... were meant to compete against Nelhalem!
a b à CPUs
July 14, 2008 1:15:23 PM

You can't increase the IPC by 10 - 25% just by upping the L3 cache from 2 to 6Mb.

Unless you radically enhanced the cache latency including the L1 and L2 ... and L3. This would require a major overhaul on the memory controller too.

Then I would be saying 5-8 % increase in performance ... still 10 % down on the latest core2 design ... their cache has also been effectively doubled.

Still ... who knows.

Someone got any more links?

a c 122 à CPUs
a b À AMD
July 14, 2008 1:21:06 PM

Reynod said:
You can't increase the IPC by 10 - 25% just by upping the L3 cache from 2 to 6Mb.

Unless you radically enhanced the cache latency including the L1 and L2 ... and L3. This would require a major overhaul on the memory controller too.

Then I would be saying 5-8 % increase in performance ... still 10 % down on the latest core2 design ... their cache has also been effectively doubled.

Still ... who knows.

Someone got any more links?


Especially with the IMC. Cache size has little effect with an IMC than does the latencies of the cache. Now for Intel a bigger cache helps a bit more but that will change with Nehalem.
a b à CPUs
July 14, 2008 1:33:04 PM

Yes ... I didn't even mention Nehalem ... I was just talking the current 45nm stuff.

Still ... if the AMD 45nm chips are good then this will be good for us all.

The Penryn 9450 / 9550/9650 quads will drop in price and that's good news for many with decent mobos with lower end dual core chips ... cheap upgrades.

!