X-bit labs

Started by: AppleJack | Date 01/09/06

Comments: 52 | Last Comment:  06/30/08

[1-20 | 21-33]

1. Intel finally did something right....65nm. Looks like 06 will be intel's year.

Apple,

They always do manufacturing right, but the P7 design is so bad it doesn't help a heck of a lot.

I agree 2006 will be a good year for Intel, with the P8 (or P6++) core coming out, and the P7 finally dying an ignoble death. The dual Yonah is already impressive, and the Merom/Conroe should be considerably better. It's hard to think of a bad scenario for Intel in the near future, but anything can happen as you no doubt know.

I really hate overclocking results put in reviews, because they are so unprofessional. No professionals I know would even consider running something out of spec. On top of this you can't guarantee a processor bought will exhibit the same frequency head room, and on top of that they are rarely apples to apples anyway. For example, why was the Athlon allowed a faster FSB? Was every attempt made to squeeze a little more clock speed out of the Presler by slightling increasing the FSB instead of the multiplier? Does anyone believe that Intel sent out a typical Presler, instead of one they know would overclock well (considering they knew this would be part of a review). AMD would do the same thing, of course, but their chip is probably less variable at this point of manufacturing.

I'm not saying it isn't worth mentioning, but to add it to benchmarks on the very top gives it grossly undue importance. As I mentioned, it is almost certainly not representative of a processor because the manufacturer will certainly cherry pick, and in any event, the vast majority of buyers would never even consider overclocking a chip costing this much, or even consider overclocking a high end processor. Even the tiny majority of people that do overclock generally buy lower clocked processors that are cheap, and then clock them to an rate that is consistent with other processors in the family that are sold at much higher cost.

It isn't useless information, because it does show AMD is pretty much out of headroom and Intel probably isn't, I just don't think it should be presented as the most important thing (which it is, being on top). I think it would be better to mention it, reiterate it at the conclusion, and not present it in such an overpowering way.

If your buying a $1000 dollar cpu I think you need to show O/C results or are you buying the $1000 cpu to browse the net LOL . Your problem is you know Intel on water will hit 5ghz so bang for the buck goes to intel . I use top of the line hardware and it cost big bucks I protect that investment with liquid cooling. Intel is coming back strong . Have you seen Conroe roadmap I have , Qt3 release Conroe EE FSB 1333 and clocked at 3.33GHZ thats going to do damage to AMD bigtime

My problem is I know Intel will hit 5 GHz? How is that my problem? It might be AMD's, if it were true, but it wouldn't bother me in the least.

So, are you saying that these processors, stock, are so slow they can only be used for surfing the web? I can surf the web with 500 MHz processors.

Overclocking is mainly for kids, and only because it is so simple now. It doesn't require any real knowledge like it used to, when you'd have to switch crystals. Professionals that need speed will NOT overclock processors or anything else for that matter. I've worked in situations where we needed every last bit of processing speed, and we'd buy the most expensive Intel sold. No one would even discuss overclocking them, you'd be look at like you were insane for even suggesting it.

So, overclocking to illustrate future headroom is legitimate, I think, and I believe it should be mentioned. But, when it becomes the focal point of the review, which it indirectly does by being mentioned so often and being placed on top, it places greater importance on it than it deserves.

Oh, and your flawed observation that I am anti-Intel shows a typical oversimplification that goes on a lot on the internet. I can criticize or give kudos to a company without being for or against them. My second paragraph should indicate pretty clearly I am not against Intel. I am not some dork that feels that a CPU maker is my family and take it personally when someone criticizes it. This type of person lacks importance in their life so they have to place it arbitrarily on, of all things, a processor maker. I am just a consumer that wants good products to choose from, so I am more than happy when either company comes out with good products. So far, I don't like either the P7 or K8 too much, so I'm excited about the P8/P6++ and I hope that it is as good as Intel says it is.

the 7 Mhz bus added to AMDs was meager in the least.

But the uneducated writer is an exaqmple of the new any idiot can overclock craze.

Higher Clock is always better. in the mind of this craze.


Higher Buses should be a priority over higher multipliers.
Even if a lower multiplier is necessary.
a 3000mhz chip with 1111bus should outperform a 3200mhz chip with 1000bus.

note im not gonna bother with the math so the numbers are illistration guides only

I think you missed the point, it was to illustrate that it was not an apples to apples comparison. Not only because it has a greater FSB, but because they didn't try to the same extent for the Intel platform. It isn't intended as something negative directed at AMD, just at the whole practice of overclocking and how it creates more questions than it answers.

Increasing FSB isn't the panacea, because the limitation is on the memory. If you have to increase timings, you aren't getting much. So, while I'd agree that memory is a huge bottleneck, I don't agree that increasing the FSB will help much because the real limitation is on the memory itself. I am sure they are doing everything they can to make memory faster, but, admittedly, it lags very far behind CPUs.

2. Dell is advertising an intel 955 oc to 4.26....so why not go 4.7 or maybe 5.0. I wonder how fast this bitch can really go??

One magazine clocked it to 4.8GHz with the Intel 975X motherboard.

3. Thanks X-Bit for the Adobe, 3D Studio Max & Maya benchmarks
It's exactly what I've been after

4. I was wondering, let's said if increas the fsb-frequency instead of the multiplier for the Intel's 955 ExtremeEdition to 4.26GHz, will the results still the same??? Since Intel's Processors love high bandwidth and with high fsb-frequency, it takes advantage from the fsb bandwidth as well.

Evogenix

5. "This was actually one of the reasons why our article has been severely criticized by many readers."

Good to see that Xbitlabs does read the comments :)

6. learn how to install a heatsink, and overclock. i know the FX-60 could do better than 2.9 GHz at a lower temperature.

7. Great review. The directly comparing between 955EE and FX60 are the most important one. However I also liked to see the overclocked numbers. After reading a lot of reviews and forums, I must say that overclocking of the high end CPUs (with air-coolers and max. +10% in voltage, on good oc-boards like DFI/Asus), are more or less allways giving you these values:

Dualcore Presler (950/955) = 4.26GHz
Singelcore Cedar Mill (3.6GHz) = 4.5GHz
Dualcore A64 (FX60) = 2.9GHz
Singelcore A64 (FX57) = 3.0Ghz

Overclocking can of course be further improved with watercooling or phasechange. However these numbers has less interest to most of us. The overclocking results tells us:

1) Intel has with 65nm frequency room enough to release a dualcore 4.0GHz and singelcore 4.2GHz, if needed before Conroe.
2) AMD will probably only release the FX62 with the next improved manufacturing F-stepping on 90nm (expected to launch in april/may with a new air-overclocking limit around 3.5GHz).
3) After overclocking with air, AMD still tops out Intels dualcores in most cases (and always in games).

Where did you get your information about the new revision based on 90 nm?

This article was pulished dec.2005 about the new 90nm manufacturing stepping:
http://www.pcworld.com/resource/article/0,aid,123850,pg,1,RSS,RSS,00.asp

Basicly it only says, that transistor performance wil be 20% better on 65nm combined with todays DSL straining and 2 new straining techniques (Embedded SiGe & Memorization). However, the new manufactoring proces will start allready on 90nm, so when the new M2-processors are released it will be on a new F-stepping.

Right now there is no info about who much better the F-stepping will be, but I hope the new way of straining will give us the 20% better overclocking allready on 90nm (but it might turn out that we have to wait for 65nm in Q4 to get the expected 3.5GHz overclock).

Very nice.

Where did you see the Intel single core maximum overclocking?

I doubt that AMD .90 will do 3.5 Ghz even at single core (that AMD is phasing out?).
I think that the new materials AMD/IBM will be using will lower power consuming but not better clock potential. However sometimes both things are related but o doubt so.

I was expecting AMD would also release the FX-59 at 3.0 Ghz with the FX60. AMD seems is phasing out single core processors (last speed jump October 2004)? Or are they waiting for the Intel P4 to hit 4.0Ghz?

Intel P4s aren't ever going to hit 4GHz...

No?

If not by better 65nm maturity/refinement (I’m sure Intel will hit high clock speeds), or
by overclock, I’m sure that if they are good, any processor (not select specimens) hit high speeds you will see.

Anandtech made this overclocking preview of the cedar mill:
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2578&p=3

I think Intel is having 2 problems regarding higher officiel frequences of the Netburst. First of all the temperature problems with the stock aircooler, only makes it possible to release a max. of 4.0GHz dualcore on 65nm. Second they wont release a very fast netburst in Q2 (even if its possible), because Conroe has to be a more competetive solution in Q3.

xx

8. Well done i have only one complaint and that is the memory timings are not apple to apple

You have used Corsair 5400 UL @ 3-2-2-8 timings and it does make a differance.

But your review is much better than AT they used memory timings 5-5-5 -15

Buy the way it is not good to tell people that 2 gigs of memory is the way to go when it really helps only 1 game . Vista is not here yet.

So it looks like if you have water cooling like myself and you O/C the Both FX60 and Intel 955 ee the clear cut smart buy is Intel .

955@ 4700 would clearly dominate all the bench marks. On water 4700 GHz for the intel is not a problem . Maybe FX with hand picked core could get 3.2Ghz. The problem is the 955 doesn't have to be handpicked as these cores are hiting 5GHz easily

9. It's unbelieveable that Netburst, a 5-year-old architecture STILL allows Intel to be competitive with AMD's best.

Yes, there is more heat and more power draw, but Netburst was not a bad way to go, all things considered. But it is dead, and hopefully the replacements (Conroe, Yonah) will be worthy competitors.

If Intel merely built an Athlon64 using their 65 nm process, they would be the undisputed kings of performance.

But the on-processor memory controller is not really a big deal - see some Yonah results versus the 3800 X2.

Competition certainly gives us consumers more and better choices!

I agree netburst does seem to be dead but why does intel roadmaps show it going to the 45n . Can you say dual cores running @ 6ghz without smiling. Intel says they have the leakage problem fixed at 45n . So until Intel stops producing P4 netburst. I ain't saying its dead yet. Really look at the results the next stepping of the 65n cpu could put netburst over the top. I believe that release is going to be april.

How do you figure P7 is good based on this? It is competing against an even older architecture (K8 is minimally changed from K7, less so than Prescott was from Northwood), and is not only slower, it uses more power, and is larger. On top of this, it is done on a smaller lithography which should give it big advantages in every department.

The P7 design sucks, and sucks bad. It is atrocious that despite a smaller manufacturing lithography that this beast is a power hungry monster and still very large. Intel deserves credit for wonderful execution in manufacturing, they alway seem to be earlier than everyone else, but the design flat out sucks.

Look at it a different way, if Intel were manufacturing the K8 on with all their expertise on 65nm, and it were being compared to a 90 nm P7, what do you think the review would look like? It would be grotesque, right?

So, this failed design by Intel is rightfully dying. It always sucked, even when it first came out Intel had to do eveything they could to make the Pentium III less competitive, just so this pig would sell. In the end, it is losing market share and Intel is widely disrespected for this horrible processor. Luckily for Intel, their superior manufacturing has made the processor less inferior than it was, but it is still a sorry piece of crap.

I have little doubt the Merom/Conroe will forever dispel any feeling that the P7 should have been extended. It looks like Intel did everything right for this processor, but it is still too soon to tell for sure. Even the Yonah is very impressive, and it appears it will be improved upon quite a bit.

Good point.

I'm a "big" fan of netburst, but after reading your post it really look bad.

I thought that the P4 prescott was bad because:
-Processor design hit a maximum clock frequency point [like P3 hit 1.0Ghz remember?].

-Thermal dissipation increased because of the 1MB cache and others.

-Because of EMT64 (I don’t know the size this is taking on the processor die, everyone when saw the P4 prescott die size was talking of a dual core design because of the increased transistor count over northwood [no way 512kb more cache would take that much]), I think even today that question remains unanswered.

-I’m not sure but I think 2 northwood cores on one die is smaller than just one prescott core.

Kaz,

I think the real problem with the Prescott was, it was designed without adequately considering issues of heat. They designed it so it could hit incredible clock speeds, but when the tire hit the road, it couldn't because it was so big and got so hot. It is kind of ironic that they created this processor to hit extremely high clock speeds, but in practice a shrunk Northwood would have done better in this regard because it wasn't as limited by thermal issues as Prescott.

This is why AMD can clock their duals so close to their single processors, they don't have the atrocious heat problems that Intel does. On 90nm, the single processor chips from Intel were much further apart from the duals, in clock speed, than AMD's were.

I think you are right about the Prescott in terms of size. If I remember correctly, it was around twice the size of the Northwood. Pretty bad waste of transistors, huh?

With regards to the Pentium III, the limitation on clock speed at .18 was actually the L2 cache, not the processor core. They did create a Celeron on .18 that ran a 1.1 GHz, but Intel typically added latency to the Celeron line to help differentiate them from the Pentium III line, so my guess is these were possible not by any improvement, but by simply increasing L2 cache latency. On .13, the Pentium III-S (the processor I most prefer), is maximally clocked at 1.4 GHz by Intel, and with 512K L2 cache the horrible memory bandwidth is slightly mitigated. I have high hopes the Merom/Conroe will make the wonderful Pentium III-S obsolete. Since it is based on the Pentium III, I think this is a very likely, as well as being very positive. I am hoping they come out with a really good processor in the 30 watt range (although certainly the high end chips will come in higher than that).

10. i found this article missleading in many ways.
Firstly the first rule of overclocking has always been.

Sacrifice the multiplier to achieve Higher Bus Speeds. This wasnt illustrated in the article and may have helped describe an Athlon X2 on M2 socket (with 667mhz Ram)

The sideaffect of this was highlighted in extreme rotation in results between the OC specimens particulay in Memory intensive Benchmarks. (Intel had its bus boosted) AMD had it's multiplier boosted


These indescrepencies added to the irrelevance of the excessive mention comparing Overclocked Processors


[quote]Unfortunately, the youngest dual-core CPUs from AMD cost much more than the competitor’s solutions, which have also migrated to a faster and more economical Presler core.[/quote]
i found this phrase a little misleading it seems to imply Intel Presler is faster and econical than AMD's low end X2s.

11. What? Only 2.9G?
It is FX-60?
Really, it is one unlocked CPU(picked from 4800+). FX is not more easy than X2 to overclock.
AMD fool us.

There is not any reason for buying it.

12. in china

13. i was unable to figure out from the review if the current motherboards for the AMD X2 cpus will be able to use the new cooling technology of the cpu by simply upgrading the bios.

What are you talking about?

C&Q tech for dual core? Wasnt that already possible?

14. I think that this chip would be a waste of money, considering that the M2 DDR2 65nm chips are right around the corner. Better just to wait, unless you got 1k to burn. Thats my 3 cents.....

15. exelent AMD is my favorite procesor, exclusive for gaming performanse

16. i wonder what amd has to show intel when the conroe come out? maybe they cam start making portable jukebox like apple for better profits then.

17. The Part related to the PC mark test, in it, it is stated that"When we overclock these CPUs, the picture changes". For the Fisrt test, not the CPU or memory test, the difference is less than 1%. So I do not know how such a statement could be made.

18. overall they ae both great cpus and i would be glad to jack these from any of my rich friends

19. test the cup's in more games

20. the best processorof the world!!!

[1-20 | 21-33]