Unbiased look at the 6xx CPus

[Sentential]

Originally posted by: christoph83

Originally posted by: geforcetony
The biggest problem I see with the 6xx series is the heat output. As a former Intel-fanboy, I can say that I did seriously look at the 775 platform before landing with a S939, and it was mainly because of heat issues. Also, with the increased latency with the extra cache, it doesn't help Intel's case any.

Thermal use is down on these chips.

Also the extra cache didn't hurt it overall. 21 out of 35 tests the cache helped or stayed even. And while most of the negatives were 1% performance hits. It picked up almost 5% more fps in doom 3. A place where it needs performace boosts. And in other tests picking up 8% and 13% in some.

Not trying to nitpick though, winchester is a good choice, I have one in my main system. But I think the cache latency complaint by a lot of people here is unfounded.

BTW Sentenial I liked your thread. I was defending this chip in another thread here trying to give a different perspective on the benchmarks, saying how close these p4's and AMD really perform only to get called an intel fanboy. It's nice to see a middle ground perspective for a change. Although more benchmarks would rule =P

A friend of mine "mattsplace" has done one on his 540J. His clocked alot better than mine and has his benches avail on his thread:

http://www.ocforums.com/showthread.php?t=369127

 

[Sentential]

As for the higher wattage count of the E0 AMDs, the current theory is because AMD plans to heavily revamp the memory controller for higher capacity/density use. Methinks thats where its going

 

[Acanthus]

Originally posted by: clarkey01

Originally posted by: Acanthus

Originally posted by: Marlin1975

Originally posted by: Acanthus

Originally posted by: clarkey01

Originally posted by: Sentential
Well it is more the issue that the newer 600 cores are within 5% of gaming of AMD which is alot better considering how bad they once were.

It only pisses me off when I see people "claiming" that a 5.2ghz Intel is somehow slower than an FX55. That is simply impossible by any account

It shouldnt piss you off. As for 5% in games, im not sure about that, athlon has a very good F math calculation or something which gives it the edge.

And if your like me, you'll know that that article was bull and not worth reading. AMD is sitting on its hands at the moment, waiting for Intel to catch up.

I think when you consider how little power the athlons use in contrast to the 6 series (granted theres a slight inprovement but nothing that major over the 5xx) its still not quite up to the design on the K8.

The K8 is also at a clockspeed wall, the thermals and power dissipation on the highest chips have increased sharply. That point is kind of moot though as AMD has topped out at a slightly higher performance plateau, giving them a comfortable lead.

Where do you get that from? Their .13 chips do 2.6Ghz RETAIL without any overclocking. The .09 are still new and will go over the 2.6, let alone the added upgrades.

Reading comprehension?

The power draw and heat dissipation of the 4000+ is much higher for the tiny speed bump over 3800+.

Revision E my freind. You will see 2.8 Ghz, it may be a little hotter buy AMD have yet to implement SS (the new so called "Breakthrough IBM/AMD thing " reported a month or two back). As it stands the FX 55 is made on a .13 micron and can hit 2.6 Ghz, so its very likely they will hit 2.8 Ghz on 90 nm process, yes there will be more heat and power consumption will be up but it'll batter a 3.8.

And why pounce on AMD for higher power consumption when Intel have the same problem.

Im not pouncing on AMD, i used the qualifier "also" to indicate AMD is at the same hurdles intel is with clockspeed.

I dont see them gaining much going to 90nm, the only reason intel gained ground was Strained Silicon and the tremendous pipeline extension.

 

[clarkey01]

Not gaining much ?

See revision E

90 nm also cuts cost, more chips per wafer

they can go higher with clockspeed, you'll see 2.8 Ghz atleast, which trounces any pentium 4.

Yeah, not worth it.

 

[Sentential]

Originally posted by: clarkey01
they can go higher with clockspeed, you'll see 2.8 Ghz atleast

Says who? Do *YOU* have physical proof of this? All that I know is that the last opteron stops at 2.6ghz @ 95W

 

[Acanthus]

Originally posted by: clarkey01
Not gaining much ?

See revision E

90 nm also cuts cost, more chips per wafer

they can go higher with clockspeed, you'll see 2.8 Ghz atleast, which trounces any pentium 4.

Yeah, not worth it.

I know it cuts cost, that doesnt make the chip any faster.

Ill believe higher clockspeed when i see it.

 

[lotus503]

Originally posted by: Sentential
Absolutly ridiclous. Has any of you *ACTUALLY* owend both are are you just gonna keep spewing fanyboy crap.

Winchesters do *NOT* clock 2.6+. On average they do 2.5 MAYBE that. Overall newcastels still are much more consistant in terms of overclocking.Second my A64 setup was plagued with issues. Abiet it was chipset derived but that point remains the same.

Was the Intel more expensive? yes. Do I regert it? Hell no. It is much more stable than my K8N could ever hope to be. Obviously you pplz dont knwo what the hell you are talking about.

A64s are indeed faster in alot of respects. I agree wholly, hwoever as a platform its buggy as hell, especially on AGP

Unlike y'all we actually keep track of member's overclocks (listed here) :http://www.ocforums.com/showthread.php?t=319185

Only way you can possible get a winchester to consistantally clock past or to 2.6 is on water or phase. It is *highly* unlikely to get that on air.

Even if you go phase you run into the sub-zero issue that 1/2 of all winchesters have. Most dont boot below 0c*

2640 **cough ***cough***

2748 at 1.55 vcore ****cough*** ***cough***

Point is I am not the minority.

I have a P4 system, I have an AMD system, I am not a fanboi of either.


Liek people here have been saying, pick what you want, base don what you use your system for.

 

[gobucks]

sentinel, winchesters do, in fact, hit 2.6GHz pretty easily, even for the 3000+. Mine hit 2.5GHz at stock vcore, now it's at 2.55GHz with only a .05V vcore bump, and I'm pretty sure it's got more in it if I got better cooling and bumped up the vcore more (currently I'm using stock HSF, btw.) The 3200+ and 3500+ can often hit 2.6GHz without even a vcore bump. If you look around at the sigs on these forums, most people overclocking these chips are hitting between 2.5-2.7GHz. Besides, even if the 3000+ were, as you suggest, to top off at 2.5GHz, that's still a 39% overclock. I don't really see how AMD could be having a clockspeed wall if their chips are overclocking so damn well. After all, a given core has a bell curve distribution, and if the low end of the usable range is 2.5-2.6GHz, when they are selling at 1.8GHz, I'd say they have their process quite under control.

As for AMD's high end chips, it's kinda hard to use them as an example for AMD hitting a clockspeed wall. Yes, they eat up a fair amount of power (less than P4 5xxs, more than 6xxs), and they are stuck around 2.5GHz for non-SS models, but they are based on the old 130nm core. I don't really know why AMD decided to do that, but I guess the decision was made that their high revanue chips should be made on the more tried-and-true 130nm process, since they didn't want to piss off people paying $500+ for their CPUs. I guess the 90nm process went way better than they had hoped, and the chips they were getting exceeded the speeds they were expecting to get. Anyways, as an overclocked 90nm 3000+, 3200+, or 3500+ will show, 90nm parts, even running at high end speeds, i.e. 4000+ and FX-55, draw much less power than the 130nm parts. Throw in process refinements of Rev E and SSDOI technology, and I don't think AMD will have any trouble breaking the 3GHZ barrier and keeping temps reasonable when they transition their high end parts to 90nm.

 

[carlosd]

Not biased, based on the benchmarks and reviews.
6xx revision, not difference in performance, much more expensive.

 

[carlosd]

Originally posted by: Sentential
Absolutly ridiclous. Has any of you *ACTUALLY* owend both are are you just gonna keep spewing fanyboy crap.

Winchesters do *NOT* clock 2.6+. On average they do 2.5 MAYBE that. Overall newcastels still are much more consistant in terms of overclocking.Second my A64 setup was plagued with issues. Abiet it was chipset derived but that point remains the same.

Was the Intel more expensive? yes. Do I regert it? Hell no. It is much more stable than my K8N could ever hope to be. Obviously you pplz dont knwo what the hell you are talking about.

A64s are indeed faster in alot of respects. I agree wholly, hwoever as a platform its buggy as hell, especially on AGP

Unlike y'all we actually keep track of member's overclocks (listed here) :http://www.ocforums.com/showthread.php?t=319185

Only way you can possible get a winchester to consistantally clock past or to 2.6 is on water or phase. It is *highly* unlikely to get that on air.

Even if you go phase you run into the sub-zero issue that 1/2 of all winchesters have. Most dont boot below 0c*

Here again the intel fanboy.

Winchesters do clock 2.6+. So, if not How the hell I have my CPU working at 2.65GHz perfectly stable with air cooling?

The A64 AMD plataform is no less stable than intel, as has been showed in this forums with stable systems working over specs (Not beacause you got stability problems with your board, it means the plataform is not stable. I mean, I have had also lots of stability problems with 775 plataforms specially beacuse of the 5xx CPUS working at the limit of is thermal desing), and in anandtech benchmarks with oveclocking guides and recomendations (And the main recommendation is not intel).

Sorry but the one with no idea of what is talking about is you.

 

[Capt Caveman]

It seems though that a good number of the newer Winchesters are not great overclockers. Most folks are only able to hit 2400-2500. There are exceptions but it seems to be a minority.

Sentential is a fanboy for neither platform. If you read his ocforums tests on both platforms using a large number of common components, the results of the 6xx aren't as bad as a number of people are spewing. He has experience from both sides, unlike the majority of us.

 

[Leper Messiah]

Originally posted by: Sentential
As for the higher wattage count of the E0 AMDs, the current theory is because AMD plans to heavily revamp the memory controller for higher capacity/density use. Methinks thats where its going

hey. There's this button. Its called EDIT!!! what was that, quadruple post?

 

[clarkey01]

Originally posted by: Sentential

Originally posted by: clarkey01
they can go higher with clockspeed, you'll see 2.8 Ghz atleast

Says who? Do *YOU* have physical proof of this? All that I know is that the last opteron stops at 2.6ghz @ 95W

Yeah I Have FAB 30 in my bedroom, hectors my uncle and sanders is my grandad.

Im 90% sure 2.8 Ghz will be achieved on the 90 nm process, not that it even has to, at 2.6 Ghz (even on 130 nm process) it still packs quite a punch. Proof ?

 

[clarkey01]

Originally posted by: gobucks
sentinel, winchesters do, in fact, hit 2.6GHz pretty easily, even for the 3000+. Mine hit 2.5GHz at stock vcore, now it's at 2.55GHz with only a .05V vcore bump, and I'm pretty sure it's got more in it if I got better cooling and bumped up the vcore more (currently I'm using stock HSF, btw.) The 3200+ and 3500+ can often hit 2.6GHz without even a vcore bump. If you look around at the sigs on these forums, most people overclocking these chips are hitting between 2.5-2.7GHz. Besides, even if the 3000+ were, as you suggest, to top off at 2.5GHz, that's still a 39% overclock. I don't really see how AMD could be having a clockspeed wall if their chips are overclocking so damn well. After all, a given core has a bell curve distribution, and if the low end of the usable range is 2.5-2.6GHz, when they are selling at 1.8GHz, I'd say they have their process quite under control.

As for AMD's high end chips, it's kinda hard to use them as an example for AMD hitting a clockspeed wall. Yes, they eat up a fair amount of power (less than P4 5xxs, more than 6xxs), and they are stuck around 2.5GHz for non-SS models, but they are based on the old 130nm core. I don't really know why AMD decided to do that, but I guess the decision was made that their high revanue chips should be made on the more tried-and-true 130nm process, since they didn't want to piss off people paying $500+ for their CPUs. I guess the 90nm process went way better than they had hoped, and the chips they were getting exceeded the speeds they were expecting to get. Anyways, as an overclocked 90nm 3000+, 3200+, or 3500+ will show, 90nm parts, even running at high end speeds, i.e. 4000+ and FX-55, draw much less power than the 130nm parts. Throw in process refinements of Rev E and SSDOI technology, and I don't think AMD will have any trouble breaking the 3GHZ barrier and keeping temps reasonable when they transition their high end parts to 90nm.

Guy knows what he's talking about, but I think they'll hit 2.8 Ghz. & stay there.Good post, you got it in one.

 

[clarkey01]

Originally posted by: Acanthus

Originally posted by: clarkey01
Not gaining much ?

See revision E

90 nm also cuts cost, more chips per wafer

they can go higher with clockspeed, you'll see 2.8 Ghz atleast, which trounces any pentium 4.

Yeah, not worth it.

I know it cuts cost, that doesnt make the chip any faster.

Ill believe higher clockspeed when i see it.

90 nm have tweaks over thier older 130 nm parts, so the 90 nm chip is 5% faster clock for clock, extra head room for Mhz and cuts cost seems an good gain for me.

 

[jbh129]

The fun with intels is that with some good cooling you can easily break 5ghz and do some things that the winchesters have trouble with (due to their subzero problems), such as 23sec SuperPI's.

If the new Nvidia chipset for Intel is a good performer then I will probably buy a 600 series because I already have a winnie and nforce4 is not an upgrade at the moment.

 

[clarkey01]

Originally posted by: jbh129
The fun with intels is that with some good cooling you can easily break 5ghz and do some things that the winchesters have trouble with (due to their subzero problems), such as 23sec SuperPI's.

If the new Nvidia chipset for Intel is a good performer then I will probably buy a 600 series because I already have a winnie and nforce4 is not an upgrade at the moment.

Iv seen an FX 55 hit 3.5 Ghz with "good cooling", 5 ghz ? what on? air ?

 

[jbh129]

air is not good cooling (I was talking phase change) and I call shenanigans if you say a FX-55 can run 3.5 on air. The world record I believe is 3820. And that is just a cpuz screen shot.

 

[clarkey01]

stable at 5 ghz ?

 

[Sentential]

Originally posted by: gobucks
sentinel, winchesters do, in fact, hit 2.6GHz pretty easily

It is about as easy as hitting 4.1ghz on a P4. It isnt easy and the ratio is about 30:70 in getting one that does. Plus for some odd reason the steppings arent getting better but more or less are the same.

 

[Sentential]

Originally posted by: carlosd

Originally posted by: Sentential
Absolutly ridiclous. Has any of you *ACTUALLY* owend both are are you just gonna keep spewing fanyboy crap.

Winchesters do *NOT* clock 2.6+. On average they do 2.5 MAYBE that. Overall newcastels still are much more consistant in terms of overclocking.Second my A64 setup was plagued with issues. Abiet it was chipset derived but that point remains the same.

Was the Intel more expensive? yes. Do I regert it? Hell no. It is much more stable than my K8N could ever hope to be. Obviously you pplz dont knwo what the hell you are talking about.

A64s are indeed faster in alot of respects. I agree wholly, hwoever as a platform its buggy as hell, especially on AGP

Unlike y'all we actually keep track of member's overclocks (listed here) :http://www.ocforums.com/showthread.php?t=319185

Only way you can possible get a winchester to consistantally clock past or to 2.6 is on water or phase. It is *highly* unlikely to get that on air.

Even if you go phase you run into the sub-zero issue that 1/2 of all winchesters have. Most dont boot below 0c*

Here again the intel fanboy.

Winchesters do clock 2.6.

They do with very little consistancy which is the *real* issue. Ive got a friend who has his to 2.76 on air. Is that typical? No is 2.65? no.

The problem is that people have a skewed view of what is "realistic". Im not saying it isnt possible, rather that it is unlikely.

Hell the problems on Intel are worse. The odds of hitting 4ghz are as hard as hitting 2.6. The issue is that people should not base a platform on how the top 2% chips clock to but rather the whole.

 

[PetNorth]

Originally posted by: gobucksAs for AMD's high end chips, it's kinda hard to use them as an example for AMD hitting a clockspeed wall. Yes, they eat up a fair amount of power (less than P4 5xxs, more than 6xxs)

No, very far of that (obviously, full load, it is what matter, who cares idle):

http://techreport.com/reviews/2005q1/pentium4-600/index.x?pg=16

 

[Sentential]

Thanks for that link... This would explain why at stock my 640 appeared to run cooler at idle/stock. According to those graphs that is correct

 

[PetNorth]

As I said, who cares idle...

 

[Sentential]

True but every little bit counts, especially when silence is prefered on idle. However the wattage difference is very minimal. Its not like the ole days where AXPs lacked C2 state. That wasnt fun :-/