AMD 45nm Deneb Pictures, CPU-Z and Super Pi Results

[0g1]

What I dont understand is, why AMD can only get HT to run at 2-2.2 Ghz and therefore their L3 Cache speed. Meanwhile Intel can run Nehalem L3 at the CPU speed ... 3.2Ghz+ ... and their L1 and L2 caches seem to have lower latency as well. Their L2 cache is lower latency than Core 2's, thus why its Super Pi scores are so good. Seems AMD's cache speeds have been the same for about 5 years, ever since A64 first came out. Meanwhile, Intel's have been improving very rapidly since the P4.

[_Lone_Wolf_]

What I dont understand is, why AMD can only get HT to run at 2-2.2 Ghz and therefore their L3 Cache speed. Meanwhile Intel can run Nehalem L3 at the CPU speed ... 3.2Ghz+ ... and their L1 and L2 caches seem to have lower latency as well. Their L2 cache is lower latency than Core 2's, thus why its Super Pi scores are so good. Seems AMD's cache speeds have been the same for about 5 years, ever since A64 first came out. Meanwhile, Intel's have been improving very rapidly since the P4.

Power budget limitations would be my guess, they seem obsessive about performance per watt over simply raw performance numbers. The SRAM cells are identical between Phenom's level 2 and 3 caches so it seems unlikely they're causing the limitation, that said I can recall JumpingJack suggesting AMD's 65nm process may have issues with excessive wire delay which would impact on level 3 cache.

I wonder if AM3 will see a boost in northbridge clockrate as an enticement to upgrade. The 1.8GHz clock in the enginering sample might be simply to maintain AM2+ compatability, or maybe thats just wishfull thinking... (?)

[Rammsteiner]

Well, the only thing he's gonna do is end up banning his ass in the end.
And I bet won't be missed.


RLY, I think AMD users have better judgement than to attack a P4 fanboy.

I think you're right.

Sod these unfortunate accidents, they just want all the attention.

Anyhow Tony managed after lots of reboots etc a very lucky 4Ghz boot into Windows with a 9850BE under water cooling. Although as he said his self, after so many reboots and then get lucky it's not even worth notifieing specially.

Og1, another reason might be the use of high/k and metal gates. I dont know exactly where they would be used in a CPU, although in the end a CPU has lots of transistors in aout every section. Anyhow, AMD doesnt use these high/k and metal gates yet and thus might not be able to run everything at higher speeds.

Could be just crappy L3 though really, no clue

[KeZzZu]

Hey anyone from Asia? could ask that dude run Cinebench!

ps: Mods could clean up this thread, way too far has gone this madness!

[Hornet331]

Power budget limitations would be my guess, they seem obsessive about performance per watt over simply raw performance numbers. The SRAM cells are identical between Phenom's level 2 and 3 caches so it seems unlikely they causing the limitation, that said I can recall JumpingJack suggesting AMD's 65nm process my have issues with excessive wire delay which would impact on level 3 cache.

I wonder if AM3 will see a boost in northbridge clockrate as an enticement to upgrade. The 1.8GHz clock in the enginering sample might be simply to maintain AM2+ compatability, or maybe thats just wishfull thinking... (?)

interesting theorie, if it really turns out that 65nm was "bad" process and 45nm is a lot better it might be possible that deneb can outperform yorkfield. I think they still can't outperfrom nehalem (with the results that we where given till now), but it seems they can demand more money for deneb then for K10 which is a good thing.

[Brother Esau]

AMD is moving forward in a very positive way especially since the acquisition of ATI but the only thing holding them back are the people making the high up decisions Perhaps the Stock Holders need to break out the BAN STICK and hit Hector over the Ass with it and throw him the hell out of the AMD Camp

[Rammsteiner]

interesting theorie, if it really turns out that 65nm was "bad" process and 45nm is a lot better it might be possible that deneb can outperform yorkfield. I think they still can't outperfrom nehalem (with the results that we where given till now), but it seems they can demand more money for deneb then for K10 which is a good thing.

I dont think anyone seriously expects Deneb to give Bloomfield a kick. However, Lynnfield might be possible though, be it neck to neck or either with a few price adjustments.

But as you stated, as long as they can start making more money with Deneb over Agena it's a start. Also Deneb FX doesnt seem to be cancelled like Agena FX (thus far), so it looks like AMD is a lot more confident about Deneb already.

But well, only we've got thus far is a C0 Super PI benchmark. Lets hope we'll see either another more usefull C0 bench or even a C1.

[Hornet331]

Also Deneb FX doesnt seem to be cancelled like Agena FX (thus far), so it looks like AMD is a lot more confident about Deneb already.

Is the FX platform still up-to-date? Or are there gona be new boards for Deneb?

But imho FX platform will play in the same ballpark as skulltrail-> it a niche in a niche market.

[Rammsteiner]

Is the FX platform still up-to-date? Or are there gona be new boards for Deneb?

But imho FX platform will play in the same ballpark as skulltrail-> it a niche in a niche market.

Hmmm, FX should be more like QX's I think.

4x4 was AMD's Skulltrail but they cancelled it.

But at this point the FX platform is pretty much non-excistant. Black Edition more or less does FX's job but at a low price. I dont know what to expect from Deneb FX though. Will the BE line get out? Or would Deneb FX feature good binned chips? More L2 cache? Ive no clue

[Hornet331]

Hmmm, FX should be more like QX's I think.

4x4 was AMD's Skulltrail but they cancelled it.

But at this point the FX platform is pretty much non-excistant. Black Edition more or less does FX's job but at a low price. I dont know what to expect from Deneb FX though. Will the BE line get out? Or would Deneb FX feature good binned chips? More L2 cache? Ive no clue

thx, good to know. Maybe amd decided to drop FX in aid of BE. I think the last FX (4x4) didn't throw a good light at that platform.

[Cooper]

If you see inappropriate post just report it. Replying to it in the same manner will only make things worse for both parties.

[awdrifter]

thx, good to know. Maybe amd decided to drop FX in aid of BE. I think the last FX (4x4) didn't throw a good light at that platform.

If they do another FX, it should be on the 4x4 platform only, and leave BE for the single socket cpus.

[Xoulz]

Nice to see AMD's 45nm CPU, even tough it is too late to really compete with Intel.

Uno, I was thinking the same thing. But is that really true...?




.

[Rammsteiner]

If they do another FX, it should be on the 4x4 platform only, and leave BE for the single socket cpus.

Since hornet brought it up a bit, Im wondering what's going to happen with either BE and FX's.

I mean, current BE's actually have the same features as we used to have from FX's, namely unlocked multipliers. But, the FX's used to sell as a high-end product around the 900~1K Euro while the BE sells currently for only 175~200 Euro. However, there's no FX CPU around at the moment.

Since the FX line is well-known to be high-end, but Intel managed to outperform it in both clocks and clock for clock performance AMD can't sell the FX line with the usual price.

So, I was thinking maybe AMD replaced the FX's by BE's and sells it way lower so when Deneb gets out, or any other better performing core, they can introduce the FX line again with the old prices?

Because, if they would sell the current BE's as FX's for the current BE's price and AMD gets Deneb/better core out everyone would go mad if the FX line would go from 200 Euro all the way to 1K Euro.

Ive no clue if that might be the theory behind it, doesnt sound stupid though.

Whether Deneb FX should be released on a 4x4 platform again... Not sure. I think both 4x4 and Skulltrail are dead. Not even because of the price, but just purely because of the very few options. There are really almost no choices regarding motherboards, and since it's such a niche market the chance of motherboard revisions (either a lay-out fix or a better chipset) is really small.

Anyhow, I do think Deneb FX should come up with a little bit more than unlocked multipliers, maybe twice the L2 cache? High/k and metal gates?

[Rammsteiner]

Uno, I was thinking the same thing. But is that really true...?

Not sure. I do see it happen Deneb is going to compete with Lynnfield. In the end you got to ask your self where it's going to compete with Intel. It won't compete with Bloomfield, Im pretty sure about that.

But Lynnfield/Yorkfield, I do see it happen. Maybe a bit hopefull, but a realistic option.

It also depends a bit how much DDR3 would influence Phenom's performance.

[Xoulz]

Oh look, the usual battle lines have already formed up. How tiring.

It's faster in superPI, that's great...for SuperPI users. As for me, I will be interested in seeing how it performs in applications I use daily relative to whatever else is available at the time.

C'mon.... why did you have to inject common sense into this thread?



.

[perkam]

C'mon.... why did you have to inject common sense into this thread? .

Get off of it Xoulz. Stick to the topic. Ty Cooper

The problem with these clocks is that we were given the same results when the 9850BE came out, before later revs started crapping out at 2.6 and 2.7Ghz.

I'm hoping the clocking issue will be solved with the shrink and that we do not have to wait for the 45nm hi-k version in H2 2009 for us to have 2.9-3.2 Ghz clocks for 24/7 use.

Perkam

[justapost]

I'm hoping the clocking issue will be solved with the shrink and that we do not have to wait for the 45nm hi-k version in H2 2009 for us to have 2.9-3.2 Ghz clocks for 24/7 use.

Last rumours i heard say high-k might come with the 32nm die shrink not earlier.
The chip does not seems to be handed out from AMD, they learned from their fiasco with the tlb issues and the early 65nm showings. Running superpi 1, at 3.4GHz does not say anything about the overall stability. We get an first impression of the cache and ipc improvement that's it for now.

[Sparky]

Last rumours i heard say high-k might come with the 32nm die shrink not earlier.
The chip does not seems to be handed out from AMD, they learned from their fiasco with the tlb issues and the early 65nm showings. Running superpi 1, at 3.4GHz does not say anything about the overall stability. We get an first impression of the cache and ipc improvement that's it for now.

This is true definitely. I ran superpi 1M on my opteron 165 at just over 3GHz, but the system was far from being totally stable. Which by the way even at that speed it only ran like 28 seconds at the lowest, so at least from a superpi perspective Deneb is showing to be better lol

[Tony]

you know I think AMD really do not care that they are slower than Intel at this time, they are building the "platform" which we all keep forgetting about. The GPU's look to be kicking ass now, chipset issues are getting ironed out and CPU clock speed and efficiency are climbing...which is what they had panned.

[metro.cl]

Did anyone see the date of the BIOS? seems weird to me that if you have access to this CPU you use a 6 month old BIOS.


Clocks look good but the vCore is quite high for those MHz.

[massman]

Metro.cl, that's not the only think that's weird ...

[Rammsteiner]

Did anyone see the date of the BIOS? seems weird to me that if you have access to this CPU you use a 6 month old BIOS.


Clocks look good but the vCore is quite high for those MHz.

Also weird the BIOS supports this CPU in the first place. I would have thought a BIOS update was needed there

[metro.cl]

Metro.cl, that's not the only think that's weird ...

What else?

Also weird the BIOS supports this CPU in the first place. I would have thought a BIOS update was needed there

I'm not sure but current bioses have some code for unreleased CPUs (not 6 month old thought) like Phenom X2 or Phenom FX, you just cant get those CPUs.

I did some normal tricks with Photoshop but images look valid.

[Newblar]

you know I think AMD really do not care that they are slower than Intel at this time, they are building the "platform" which we all keep forgetting about. The GPU's look to be kicking ass now, chipset issues are getting ironed out and CPU clock speed and efficiency are climbing...which is what they had panned.

yeah, i think so too. the new integrated gpu's and chipsets seem to be outperforming their intel counterparts