Phenom B3 launches next week

[coldpower27]

Originally posted by: Extelleron

Originally posted by: Idontcare

Originally posted by: Extelleron
From looking at the article, it looks to me like they are either running EE versions of the 90nm CPUs or undervolting them, because stock X2 5000+ 90nm voltage is not 1.184V. According to newegg, stock voltage should be 1.30/1.35V: http://www.newegg.com/product/...turesmx-_-NA-_-NA-_-NA

The reason that you see such power consumption in the THG article is that the 90nm F3 is running at 1.184V, while the F2 90nm is running at an unknown voltage, since the load CPU-Z image for the F2 CPU is the same image as for the F3 CPU. For the 65nm G1, voltage is 1.328V and for G2, voltage is 1.344V

So all this test shows is that if you undervolt a 90nm F3, you can acheive the same results as a stock 65nm G2.

Thanks for bringing up the voltage...I did a double take when I read the article and saw the load volts in the CPU-Z shots, but then having no recent experience with the AMD side I figured that was something they intentionally did with the 65nm X2's to boost the number of chips that fell into the speedbin.

(same TDP for 90nm and 65nm because the 65nm voltages are higher, but because of this there are even more chips/wafer that hit that speedbin at 65nm)

So is it really just fubar'ed? Do the G2's really not require 1.34'ish volts to hit their speeds? Or are the 90nm numbers just BS because they are using EE chips?

Really to make the analyses more robust they should have intentionally undervolted all the chips so they were functioning at their minimum stable voltage for stock clockspeeds while fully loaded. Then you'd have a clearer assessment of the underlying technology. Not perfect, but it would be better.

The voltage for the 65nm parts seems to be pretty accurate - AMD's Brisbane CPUs have a fairly high stock voltage, looking at newegg, voltage for 5000+ Brisbane is 1.325-1.375V. Stock voltage for 5000+ BE, guaranteed G2, seems to be a bit lower @ 1.25-1.35V.

AMD does have a 90nm 2.6GHz part with a 65W TDP - the Opteron 1218 HE is 2.6GHz w/ 65W @ 1.20-1.25V. From what I see though, there is no X2 5000+ 90nm that ran at 65W.... IDK, but whatever, THG results are clearly FUD.

http://products.amd.com/en-us/...opCPUDetail.aspx?id=45

The 5000+ 90nm 65W did exist, I think the highest AMD got on 90nm with 65W was 5200+
at 2.6GHZ.

The tweaked the hell out of 90nm though. It wasn't until the Brisbane 5200+ where AMD finially surpassed the 90nm process in terms of clocks within the same thermal envelope.

 

[coldpower27]

Originally posted by: Kuzi
You see AMD's 90 nm process was excellent, throughout it's life time AMD kept on tweaking and improving it. They had the time and resources to do that because Intel were still dealing with their Netburst mess.

After the release of 65nm Core 2 processors everything changed, Intel got a huge boost in performance over X2s and also were a process generation ahead (smaller, cheaper, less heat and power).

AMD released 65nm processors about 18 months after Intel (that is normal for AMD), but the clocks went down instead of going up. I mean if the fastest Athlon X2 at 90nm clock to 3.2Ghz, you would expect the new process generation to clock at least 3.6Ghz or so. Instead Brisbanes only ran at 2.6GHz-2.8Ghz.

Phenom being a quad core would have a hard time just getting to 2.6GHz, and if you look at Phenom OC results after increasing the frequency a few hundred Mhz, the power draw becomes unbelievably high, probably the cause of a serious leakage problem.

It remains to be seen if/how AMD will fix this leakage problem with 45nm, I do have high hopes for Shanghai, at least to allow AMD to be competitive again. A 10-15% clock per clock performance increase over Phenom and at least 3GHz frequency would help them do that. But can they do it?

Do keep in mind, that like the 65nm process the 90nm process started out with lower clocks at first with the Winchester cores vs the Newcastles, it wasn't till San Diego that AMD got higher clocks out of 90nm over the previous 130nm generation. But unfortunately for AMD it looks like it is simply taking too long for 65nm to mature in terms of speed. Were below the point of where we should be if using the 130-90 transistion as a guideline.

Also keep in mind the 90nm process was pretty much 3 years old when AMD extracted 3.2GHZ out of it.

 

[CTho9305]

coldpower27, can you extrapolate to where the Brisbanes would be frequency-wise if there were 89W or higher bins? Putting your 2 posts together it looks like 90nm did 2.6GHz at 65W and 65nm does 2.8GHz?

 

[dennilfloss]

So how does that new CPU compares to an X2 4200+@2.2GHz in terms of its default performance for those of us who rarely overclock?

 

[Sylvanas]

Originally posted by: dennilfloss
So how does that new CPU compares to an X2 4200+@2.2GHz in terms of its default performance for those of us who rarely overclock?

A Phenom B3 would be a worthy upgrade over a 4200+ stock for stock.

 

[Extelleron]

Originally posted by: Sylvanas

Originally posted by: dennilfloss
So how does that new CPU compares to an X2 4200+@2.2GHz in terms of its default performance for those of us who rarely overclock?

A Phenom B3 would be a worthy upgrade over a 4200+ stock for stock.

Phenom should be around 20% faster clock for clock in most situations, plus 4 cores instead of 2, and depending on which model, higher frequency as well.

Pretty big upgrade. In multithreaded situations, you could expect 2X+ performance.

 

[Kuzi]

Originally posted by: coldpower27
But unfortunately for AMD it looks like it is simply taking too long for 65nm to mature in terms of speed. Were below the point of where we should be if using the 130-90 transistion as a guideline.

Also keep in mind the 90nm process was pretty much 3 years old when AMD extracted 3.2GHZ out of it.

Right now AMD does not have the time/resources/ or money to waste on their 65nm process. Intel will be flooding the market with cheap and fast 45nm processors in the 2nd half of this year.

I don't think they will even try improving 65nm Phenoms much beyond B3, but they are probably working hard on 45nm to release Shanghai as soon as possible. They had running Shanghai CPUs since January, so there is a chance they can release it towards the end of this year. Still it is 4 months behind Nehalem that Intel demonstrated in September.

I've been waiting since last summer to build a new AMD system, but at the current state it's not worth it yet.

 

[byronm]

Originally posted by: Kuzi
I've been waiting since last summer to build a new AMD system, but at the current state it's not worth it yet.

Its a shame AMD wasn't able to get the phenoms out 2 years ago when they were fabbed. I'm curious as to why they sat on them and if the merger of ATi is what gave away the large CPU market share ffor the short term.

Phenoms are a great choice for AM2 users to get quad core cheaper than a Q6600 because a Q6600+MB costs more than a CPU alone however even AMD didn't get much support from the existing am2 motherboard manufactures since they're only motivated to sell the next gen.

So in hind sight, the phenoms should have been AM2+ ONLY, pushed to market much earlier and they should have squashed the errata fears much earlier in the consumer market than what happened.

Happy Phenom customer here (but anxiously awaiting a new "beast" to try out)

 

[v8envy]

Also remember that on an AM2 board the Phenom is a less stellar performer than on an AM2+ board. I remember seeing a review where the performance haircut was as much as 50% over the same CPU in an AM2+ board.

Of course board makers want to sell you new product. Supporting old product doesn't bring in revenue. So from that standpoint you should be comparing new board + new CPU + 533 mhz ram vs board + cpu + 333 or 400 mhz ram. Reviews where the Phenom keeps up with a stock clock Q6600 are done with the pricier PC8500 RAM. Pair the Phenom with the cheap stuff and it takes another hit -- the core2 not so much.

 

[nyker96]

Guess they are having problem upping clocks on B3s.

 

[Kuzi]

Originally posted by: Kuzi
Right now AMD does not have the time/resources/ or money to waste on their 65nm process. Intel will be flooding the market with cheap and fast 45nm processors in the 2nd half of this year.

I don't think they will even try improving 65nm Phenoms much beyond B3, but they are probably working hard on 45nm to release Shanghai as soon as possible. They had running Shanghai CPUs since January, so there is a chance they can release it towards the end of this year. Still it is 4 months behind Nehalem that Intel demonstrated in September.

From Anand's Phenom B3 article:
"AMD is still on track to begin shipping its first 45nm Phenom processors (Deneb core) by the end of this year and it doesn't make sense to waste time and resources respinning a 65nm Phenom, when presumably these clock speed issues are addressed at 45nm."

What I was thinking too, not worth it to improve B3 more than that.

Notice the Phenom X4 9850 Black Edition at 2.5GHz had good performance overall (at $235), I only wish AMD clocked the IMC at 2.5GHz too instead of 2GHz. I mean it's a Black Edition for overclockers, who cares if it will draw a bit more power/heat.

BTW, nice article Anand, just when I was thinking the quality of the articles here were not like how they used to be, we get a pretty good one

 

[Kuzi]

Originally posted by: Kuzi

Originally posted by: Kuzi
Right now AMD does not have the time/resources/ or money to waste on their 65nm process. Intel will be flooding the market with cheap and fast 45nm processors in the 2nd half of this year.

I don't think they will even try improving 65nm Phenoms much beyond B3, but they are probably working hard on 45nm to release Shanghai as soon as possible. They had running Shanghai CPUs since January, so there is a chance they can release it towards the end of this year. Still it is 4 months behind Nehalem that Intel demonstrated in September.

From Anand's Phenom B3 article:
"AMD is still on track to begin shipping its first 45nm Phenom processors (Deneb core) by the end of this year and it doesn't make sense to waste time and resources respinning a 65nm Phenom, when presumably these clock speed issues are addressed at 45nm."

What I was thinking too, not worth it to improve B3 more than that.

Notice the Phenom X4 9850 Black Edition at 2.5GHz had good performance overall (at $235), I only wish AMD also clocked the IMC at 2.5GHz instead of 2GHz. I mean it's a Black Edition for overclockers, who cares if it will draw a bit more power/heat. And OC limit still seems around 2.8-3GHz.

BTW, nice article Anand, just when I was thinking the quality of the articles here were not like how they used to be, we get a pretty good one

 

[Martimus]

Originally posted by: Kuzi

Originally posted by: Kuzi
Right now AMD does not have the time/resources/ or money to waste on their 65nm process. Intel will be flooding the market with cheap and fast 45nm processors in the 2nd half of this year.

I don't think they will even try improving 65nm Phenoms much beyond B3, but they are probably working hard on 45nm to release Shanghai as soon as possible. They had running Shanghai CPUs since January, so there is a chance they can release it towards the end of this year. Still it is 4 months behind Nehalem that Intel demonstrated in September.

From Anand's Phenom B3 article:
"AMD is still on track to begin shipping its first 45nm Phenom processors (Deneb core) by the end of this year and it doesn't make sense to waste time and resources respinning a 65nm Phenom, when presumably these clock speed issues are addressed at 45nm."

What I was thinking too, not worth it to improve B3 more than that.

Notice the Phenom X4 9850 Black Edition at 2.5GHz had good performance overall (at $235), I only wish AMD clocked the IMC at 2.5GHz too instead of 2GHz. I mean it's a Black Edition for overclockers, who cares if it will draw a bit more power/heat.

BTW, nice article Anand, just when I was thinking the quality of the articles here were not like how they used to be, we get a pretty good one

I agree. That article was very good. While I still won't buy an AMD set-up again until they are a little more competitive (maybe the 45nm shrink), but that article made a lot of sense, and showed that AMD is much closer than anyone gives them credit for.

 

[Kuzi]

ops sorry double post, I meant to just edit it.