4x4 in action...

[Viditor]

Originally posted by: harpoon84

And yet Kentsfield overclocks to 3.4 - 3.5GHz on air in most reviews. How do you explain that? Intel cherry picks good chips for reviews? XS people are getting good overclocks out of ES Kentsfields too, maybe Intel cherry picks those ES chips as well to gain good publicity?

Seriously though, wattage only becomes an issue with overclocking when it can no longer be sufficiently cooled. I'd fathom a guess that a Kentsfield @ 3.5GHz with a voltage bump would be pushing the 200W mark - probably the limits of effective and quiet air cooling on the best HSFs. Scythe Infinity anyone? If you can afford a $1000 CPU a $50 HSF is peanuts, really.

Of course they were cherry picked...that's pretty much standard practice for both AMD and Intel on their top chips for review. But that really isn't the point...
I was pointing out exactly what Anand said in his review about overclocking the Kentsfield:

"You lose some overclocking headroom given the added heat output of the extra die on the chip, not to mention that both die have to be capable of running at the overclocked speed"

Even though you can attain the higher clocks, remember that throttling kicks in much faster due to the higher heat. So, even though the clockspeed can go much higher, that will actually slow things down in terms of performance (unless you can use some exotic cooling) because of the throttling.
This is much less true for the C2D dual cores as the heat is much less...

 

[HurleyBird]

Originally posted by: harpoon84

Originally posted by: HurleyBird

Originally posted by: Skott
I'm at work and cant read the Inquirer. Our network blocks it all the time for some reason. LOL If they can price it considerably cheaper than Intel's Quad then they can rub Intel's nose a bit. The upgrade path is interesting but who's to say Intel cant do the same with their Quads? I mean if AMD can do it why cant Intel?

Are you kidding? Eight cores on one FSB would be a joke.

Depends on the speed of the FSB.

Currently, Kentsfield uses a 1066MHz FSB, or 266MHz per core. If Intel were to keep this ratio, Octo core would require a 2133MHz FSB. They would need to design a new chipset that can sustain such speeds (currently, only the very best 965P boards can hit such speeds).

I should mention that Intel can also do multiple FSBs on a single NB (ala Tigerton) - not sure if they're gonna bother for the desktop though, as the chipset would cost significantly more. Then again, so would a chipset that can handle 2GHz+...

Yeah, they can do dual FSBs, but what the guy was asking was "The upgrade path is interesting but who's to say Intel cant do the same with their Quads?" which implies the same platform as kentsfield. I don't think Intel could up the bus speed much either, it gets harder to do as you add more cores (remember how the old xeon FSB that lagged behind P4?)

 

[harpoon84]

Yeah, I'll agree on that point.

I also don't think Intel can do Octo core at this stage because AFAIK Yorkfield is two Wolfdales 'glued together', as Kentsfield is to Conroe. It might be possible if it was native quad core, glued together for 'Octo core'.

 

[harpoon84]

Originally posted by: Viditor

Originally posted by: harpoon84

And yet Kentsfield overclocks to 3.4 - 3.5GHz on air in most reviews. How do you explain that? Intel cherry picks good chips for reviews? XS people are getting good overclocks out of ES Kentsfields too, maybe Intel cherry picks those ES chips as well to gain good publicity?

Seriously though, wattage only becomes an issue with overclocking when it can no longer be sufficiently cooled. I'd fathom a guess that a Kentsfield @ 3.5GHz with a voltage bump would be pushing the 200W mark - probably the limits of effective and quiet air cooling on the best HSFs. Scythe Infinity anyone? If you can afford a $1000 CPU a $50 HSF is peanuts, really.

Of course they were cherry picked...that's pretty much standard practice for both AMD and Intel on their top chips for review. But that really isn't the point...
I was pointing out exactly what Anand said in his review about overclocking the Kentsfield:

"You lose some overclocking headroom given the added heat output of the extra die on the chip, not to mention that both die have to be capable of running at the overclocked speed"

Even though you can attain the higher clocks, remember that throttling kicks in much faster due to the higher heat. So, even though the clockspeed can go much higher, that will actually slow things down in terms of performance (unless you can use some exotic cooling) because of the throttling.
This is much less true for the C2D dual cores as the heat is much less...

Unless you can prove that those 3.5GHz overclocks touted by review sites are actually throttling, I don't think it's very relevant to the discussion. HSFs have come a long way, and something like a Scythe Infinity would handle anything Kentsfield can throw at it.

Of course, the more cores, the higher chance of a 'dud' core that clocks worse than the others. But since they are from the same wafer, the headroom should be quite similar on all cores. Some variation is expected of course but I don't expect one core to do 3GHz and the rest 3.5GHz.

It's a given that quad cores will overclock slightly worse than dual cores, but it won't be a day and night difference.

As a point of comparison, Anandtech's X6800 sample reached 4GHz on air, while the E6700 reached 3.9GHz.

So theres about a 500MHz difference in headroom between the highest binned dual cores and quad core.

 

[Viditor]

Originally posted by: harpoon84

Originally posted by: Viditor

Originally posted by: harpoon84

And yet Kentsfield overclocks to 3.4 - 3.5GHz on air in most reviews. How do you explain that? Intel cherry picks good chips for reviews? XS people are getting good overclocks out of ES Kentsfields too, maybe Intel cherry picks those ES chips as well to gain good publicity?

Seriously though, wattage only becomes an issue with overclocking when it can no longer be sufficiently cooled. I'd fathom a guess that a Kentsfield @ 3.5GHz with a voltage bump would be pushing the 200W mark - probably the limits of effective and quiet air cooling on the best HSFs. Scythe Infinity anyone? If you can afford a $1000 CPU a $50 HSF is peanuts, really.

Of course they were cherry picked...that's pretty much standard practice for both AMD and Intel on their top chips for review. But that really isn't the point...
I was pointing out exactly what Anand said in his review about overclocking the Kentsfield:

"You lose some overclocking headroom given the added heat output of the extra die on the chip, not to mention that both die have to be capable of running at the overclocked speed"

Even though you can attain the higher clocks, remember that throttling kicks in much faster due to the higher heat. So, even though the clockspeed can go much higher, that will actually slow things down in terms of performance (unless you can use some exotic cooling) because of the throttling.
This is much less true for the C2D dual cores as the heat is much less...

Unless you can prove that those 3.5GHz overclocks touted by review sites are actually throttling, I don't think it's very relevant to the discussion. HSFs have come a long way, and something like a Scythe Infinity would handle anything Kentsfield can throw at it.

Of course, the more cores, the higher chance of a 'dud' core that clocks worse than the others. But since they are from the same wafer, the headroom should be quite similar on all cores. Some variation is expected of course but I don't expect one core to do 3GHz and the rest 3.5GHz.

It's a given that quad cores will overclock slightly worse than dual cores, but it won't be a day and night difference.

As a point of comparison, Anandtech's X6800 sample reached 4GHz on air, while the E6700 reached 3.9GHz.

So theres about a 500MHz difference in headroom between the highest binned dual cores and quad core.

In reverse order...firstly, Anand's article said:

"We managed to get our QX6700 sample up to 3.2GHz (12 x 266MHz) at the CPU's stock voltage of 1.35V, which isn't bad at all considering we didn't employ any exotic cooling. Bumping the core voltage up to 1.3875V we were able to gain an extra 266MHz and run at 3.46GHz"

I didn't see the 4GHz or 3.9GHz references (could you point them out for me?)...

As to "proving" throttling, obviously it's different on every piece of silicon, but (as I haven't read all of the reviews) if you could link to a review where they actually did a full benchmark suite @3.5GHz on Kentsfield, we can compare results...

 

[harpoon84]

Yeah no problem, I was referring to Anandtech's earlier X6800/E6700 review as a point of comparison between dual core and quad core overclocking:
http://anandtech.com/cpuchipsets/showdoc.aspx?i=2795&p=18

As for throttling, I don't believe a Kentsfield @ 3.5GHz with a high end HSF will be throttling at all. It takes a very high temperature for the CPU to throttle so as long as the voltages are kept reasonable, there should be no reason it would throttle. Obviously if you try for 4GHz with 1.5V that would be a different story, but I'm talking overclocking within the bounds of current air cooling solutions.

This review goes into the overclocking aspects of Kentsfield quite thoroughly, and no mention of throttling even when 4 cores are fully loaded:
http://www.overclockers.com.au/article.php?id=519919&P=5

For the record, OCAU achieved 3.7GHz @ 1.45V on air cooling, and the author believes 4GHz would be possible on high end watercooling or phase change cooling.

 

[tshen83]

buy a Mac Pro with 8 cores coming out in a few weeks. It has dual quad core Kentsfields. Did I mention the Mac Pro has 4 16x PCI-E slots? Plus you can get intel core 2 duo(quad) chips from apple a lot cheaper because of distribution deals apple has with intel. And a mac now is a PC

 

[Viditor]

Originally posted by: harpoon84
Yeah no problem, I was referring to Anandtech's earlier X6800/E6700 review as a point of comparison between dual core and quad core overclocking:
http://anandtech.com/cpuchipsets/showdoc.aspx?i=2795&p=18

As for throttling, I don't believe a Kentsfield @ 3.5GHz with a high end HSF will be throttling at all. It takes a very high temperature for the CPU to throttle so as long as the voltages are kept reasonable, there should be no reason it would throttle. Obviously if you try for 4GHz with 1.5V that would be a different story, but I'm talking overclocking within the bounds of current air cooling solutions.

This review goes into the overclocking aspects of Kentsfield quite thoroughly, and no mention of throttling even when 4 cores are fully loaded:
http://www.overclockers.com.au/article.php?id=519919&P=5

For the record, OCAU achieved 3.7GHz @ 1.45V on air cooling, and the author believes 4GHz would be possible on high end watercooling or phase change cooling.

Reaching higher clocks says nothing about performance at those clocks...throttling occurs based on the spot internal temps (not the measured external temp), which is actually one of Intel's best features. While the Big Typhoon HSF certainly helps the chip maintain stability (as opposed to the 70C+ @3GHz of the stock HSF), please note that nobody (at least that I've seen) has measured how the chip actually performs at those levels...there's probably a good reason for that.
I do remember that the Netburst chips were often throttling at 60-65C...

As to the lack of throttling being mentioned, it's not always detectable...that's why I asked about benchmarks at overclock. I think the article was just looking for whether it was stable, not whether it actually performed well.

 

[Viditor]

Originally posted by: tshen83
buy a Mac Pro with 8 cores coming out in a few weeks. It has dual quad core Kentsfields. Did I mention the Mac Pro has 4 16x PCI-E slots? Plus you can get intel core 2 duo(quad) chips from apple a lot cheaper because of distribution deals apple has with intel. And a mac now is a PC

I think you mean Cloverton, not Kentsfield (Cloverton is the Xeon equivalent, or basically 2 Woodcrest chips glued together in an MCM). The Kentsfield cannot work with 2 CPUs on 1 board (called DP processers).
There is a big question as to how well these will perform...remember that (especially at 8 cores) the platform is a HUGE part of performance for these chips.

 

[HurleyBird]

Originally posted by: tshen83
buy a Mac Pro with 8 cores coming out in a few weeks. It has dual quad core Kentsfields. Did I mention the Mac Pro has 4 16x PCI-E slots? Plus you can get intel core 2 duo(quad) chips from apple a lot cheaper because of distribution deals apple has with intel. And a mac now is a PC

whats the point of four x16 lanes if there is no SLI/Crossfire support? Or has that situation been rectified?

Also, FBDIMM pretty much slows everything down a few speed grades. your 2.66GHz quad core would run like a 2.2GHz machine on single threaded apps--yuck.

 

[harpoon84]

Originally posted by: Viditor

Originally posted by: harpoon84
Yeah no problem, I was referring to Anandtech's earlier X6800/E6700 review as a point of comparison between dual core and quad core overclocking:
http://anandtech.com/cpuchipsets/showdoc.aspx?i=2795&p=18

As for throttling, I don't believe a Kentsfield @ 3.5GHz with a high end HSF will be throttling at all. It takes a very high temperature for the CPU to throttle so as long as the voltages are kept reasonable, there should be no reason it would throttle. Obviously if you try for 4GHz with 1.5V that would be a different story, but I'm talking overclocking within the bounds of current air cooling solutions.

This review goes into the overclocking aspects of Kentsfield quite thoroughly, and no mention of throttling even when 4 cores are fully loaded:
http://www.overclockers.com.au/article.php?id=519919&P=5

For the record, OCAU achieved 3.7GHz @ 1.45V on air cooling, and the author believes 4GHz would be possible on high end watercooling or phase change cooling.

Reaching higher clocks says nothing about performance at those clocks...throttling occurs based on the spot internal temps (not the measured external temp), which is actually one of Intel's best features. While the Big Typhoon HSF certainly helps the chip maintain stability (as opposed to the 70C+ @3GHz of the stock HSF), please note that nobody (at least that I've seen) has measured how the chip actually performs at those levels...there's probably a good reason for that.
I do remember that the Netburst chips were often throttling at 60-65C...

As to the lack of throttling being mentioned, it's not always detectable...that's why I asked about benchmarks at overclock. I think the article was just looking for whether it was stable, not whether it actually performed well.

Here's something interesting for ya Viditor:
http://www.theinquirer.net/default.aspx?article=35506

TheINQ try to 'torture' Kentsfield by running it in a 35C room with the stock heatsink - urgh! It *barely* made it through at stock speeds...

With the air con on and a room temp of 23C, he reached 3.2GHz on the stock HSF.

There are benches showing 2.66GHz vs 3.2GHz, and to me there doesn't seem to be any throttling going on.