955EE Review Finally Out

[ltcommander_data]

The 955EE is finally released and as expected it outperforms the X2 4800+ in most applications except games. What is most significant though is the overclocking potential. With the multipliers unlocked, the 955EE can easily overclock to 4GHz on the standard Intel HSF. Raising the VCore slightly from 1.3375V to 1.375V yields stability at 4.26GHz again with the standard Intel HSF.

http://www.xbitlabs.com/articles/cpu/display/presler_9.html

At 4.26GHz the 955EE dominates most benchmarks except games where its only beaten by the FX-57. It's nice to know that with new multipliers down to 12x the 955EE can use Speedstep to save a bit of power although the Netburst architecture is inheirently inefficient.

The FX-60 will be out in a few weeks though which should regain the performance lead for AMD although it'll be 20% more expensive. The X2 5000+ will be out some time after that.

Its nice that the 950D offers quite good performance compared to the previous 840EE yet is much cheaper. With the 950D price competitively with the X2 4600+ it'd be interesting to see how they compare.

 

[pat]

As for the thermal mode, the maximum temperature of the Pentium Extreme Edition 955 processor overclocked to 4.26GHz (under full workload) reached 79o C. When we measured the CPU temperature in the nominal mode, that is at 3.46GHz, it never went over 65o C. So, this indicates clearly that further frequency increase will require more efficient cooling solution onboard.

Great!! Maybe a CTX design too... Since ATX won't be adequate and BTX was a flop...

Why don't they just make cool CPU at first? I mean, something that run in the 30 degrees for normal task and in the 40 for full load ad in the 50-60 OCed....

However, we shouldn’t also disregard some drawbacks of the NetBurst architecture, which remained even after the transition to 65nm process. I am talking about high heat dissipation and power consumption of the new Presler based processors. Although more advanced production technology allowed to bring these parameters down a little bit, they still remained as high as those for the top-end single-core Prescott based CPUs. The typical heat dissipation of the new Pentium Extreme Edition 955 is even set at 130W by default.

All right. Not to bash Intel here, but I was expecting something better from the 65 nm transition..

But it is a fast CPU. But it come to a price. If at full load it got 65 degrees, that means that in the havnd of some inexperimented builders, that would mean more like 75-80..
Hey, some are able to get a Venice 3000+ running a 45 idle... While normally, it runs @ about 30 degrees.

And finally.. when Intel will finnaly stop to try to be faster and take the time to put a decent product on the market? Just put something that can run as cool as a 3200+ and as fast at a competitive price and they'll have a winner. Add dual core to that, they'll have 2 winners.

Looks like I'll stay with AMD for a while ...

 

[endyen]

as expected it outperforms the X2 4800+ in most applications except games.

Over all, the 4800 wins 10 to 8. Without games, intel 8, 4800 5.
Not what I would consider "most applications".
It becomes even more questionable when you include bapco, adobe, and 3Dmark (how the help can a chip loose in 3Dmark, yet shut the opposition out in games) Using Intel approved benches didn't help much.

At 4.26GHz the 955EE dominates most benchmarks

If you dont mind buying a new mobo every 6 months. These suckers may be "more energy efficient" than the old dual cores, but they still draw 60% more power than the X2. Lends a whole new meaning to the term sweat shop.

 

[ltcommander_data]

The X-Bit article actually didn't include 3DMark. HEXUS did do a comparison with 3DMark and the X2 4800+ scores 7377 with the 955EE scoring 7154. HEXUS probably lucky and got a better chip since they were able to overclock using multipliers to 4.26GHz without increasing VCore and using the standard Intel HSF. The 3DMark score at 4.26GHz is 7531. In real games, HEXUS found the X2 4800+ outperforming the 955EE by around 10fps.

http://www.hexus.net/content/item.php?item=4210&page=8

It seems that a new motherboard isn't required for the 955EE as both HEXUS and X-Bit found that it works fine with the i955X. Supposedly, most boards that currently support dual cores will be able to run the 955EE with only a BIOS update.

HEXUS is probably right that the 955EE should have launched as a 3.73GHz part. Since those speeds are easily reached it wouldn't effect their yields but would have provided more value. With a 3.73GHz EE pricing at $999 now, getting 2 for the same price would have been so much better. A 3.6GHz 960D for the price of the 950D would have also been nice. It seems Intel doesn't want to have Presler perform too well to prevent it from taking some of Conroe's shine. The woes of marketing.

 

[endyen]

Yes, that's right, it should say futuremark's pcmark, rather than futuremark's 3Dmark. I'd change it, but then your post wouldn't make sense. They are bothe supposed to be gamers benchmarks.
Did you see the chart on power usage? The vrms are dealing with 156 watts. With the curve on power for speed being what it is for netburst, any OC is going to reduce board life. They didn't release at 3.76 for XE and 3.6 for 950 because they cant afford to pay for new boards for everyone, every few months.

 

[ltcommander_data]

Even with the power curve of Netburst I doubt that going to 3.73GHz would have caused a sudden influx of melted motherboards. Both the i945 and i955 motherboards are designed to handle the 840EE which X-Bit found to have a power draw of 179W under full load. The 955EE draws 156W which is 23W or 13% less. Going to 3.73GHz would have increased power draw but I doubt it would jump 13% for one multiplier. The i975 motherboard would have been designed as well, if not better than the i955 motherboard. Besides, motherboards using the i955 and i975 chipsets would be designed with overclocking in mind since that is their target market giving a 3.73GHz 955EE more room for stable operation. (955EE's compatibility with the i955X chipset is fairly certain, btw, since X-Bit actually ran their entire review on the ASUS ASUS P5WD2 Premium. How much using the i955X vs. the i975X effected performance is debatable although they did use higher latency 4-4-4-12 DDR2 RAM rather than some nicer Corsair 3-3-2-8 DDR2 RAM to compete with the 2-3-2-10 DDR.)

As well, I would think there would be even less concerns about a 3.6GHz 960D Presler. A 960D would be clocked 133MHz higher than the 955EE but would use a 800MHz FSB instead of 1066MHz and wouldn't have HTT support, both of which would cut power consumption and heat. A 960D would probably be more energy efficient and cooler than the 955EE which means that it would work fine on the i945, i955 and i975.

 

[endyen]

In case you hadn't noticed, the spec for these chips is 130 watts. Now I'm sure that Asus and the boys left themselves a little head room, but getting close to 50 watts is a little scary. As to 113% of 156 being close to 179, yea, it is, but leakage current does not rise mathmaticly, but rather exponentially. I would expect the wattage of the dual cores to be 190 at least, at 3.73.
No matter how you look at it, it's a chip that will put out over 200 watts of heat, when the psu loses are included.
All this for a chip that will look clearly second rate before it arrives, since the FX60 will be available first.
Not that the 60 is cost effective. What is truly sad is that the X2 4800 is selling now for $200 less than what intel is asking for 1000 chip lots on this new and almost as good chip.

That got me thinking a little. What will $999 buy in the Amd world?
X2 A8R-MVP and A64 3700, with $277 in change, to buy ram and hdds. And yes, we are talking two A64 3700 systems, for the price of that intel XE setup.
Since I am currently running two A64 3200 systems side by each, it's not worth the upgrade to me.

 

[ltcommander_data]

Intel's TDP ratings are for 75% workload, which of course makes their real power consumption worse, but it also means that motherboards are designed stronger. The 840EE's real power consumption is 179W which means at 75% it is only 4W above the TDP of 130W, which is perfectly acceptable (from a design perspective). Now if we were to assume your value of 190W for full power consumption of a 3.73GHz 955EE is accurate, its TDP at 75% would be 143W or about 10% over spec.

However, its doubtful that the consumption would actually be that high. Now Tom's Hardware may be questionable, they did show that the system power consumption of a 3.73GHz 955EE at full load is only 5W above the standard 3.46GHz. Both of which are less than the 840EE. A major power jump doesn't occur until the 955EE transitions from 4GHz to 4.26GHz. While a 5W increase seems low, it makes sense that rapid power and thermal increases don't occur until 4.26GHz since that is the point where voltage increases are needed to increase clock speed. To limit the thermal or power increases of going to 3.73GHz, Intel could easily bin a bit tighter as HEXUS received a better 955EE that could clock to 4.26GHz with no additional voltage.

http://www.tomshardware.com/2005/12..._into_double_core_extreme_edition/page14.html

As well, the 955EE and their motherboards are designed enthusiasts and overclocking and ASUS for one has taken to include 8-phase voltage regulators with their boards that support the 955EE. 8-phase voltage regulators allow the P5WDG2-WS, the P5N32-SLI, and most likely the P5WD2-E to reduce power consumption by 10% compared to 4-phase regulators. They allow the 840EE to run 5C cooler as well. Since the 955EE's power consumption at full load of 156W is 13W lower than the 75% spec TDP of 130W, it has no problems running on standard 4-phase voltage regulator motherboards. However, the 8-phase voltage regulator ensures a 3.73GHz 955EE would have no problem running on motherboards as the 10% reduction in power consumption counteracts 10% over TDP worst-case scenario of 190W. I'm not sure about Gigabyte's current high-end boards, but they have previously featured 6-phase voltage regulators providing a 5% power reduction which with design tolerances would likewise be acceptable if a 3.73GHz 955EE really goes up to 190W.

http://www.anandtech.com/mb/showdoc.aspx?i=2580