Sign in with
Sign up | Sign in
Your question

"Real" Conroe Benches

Last response: in CPUs
Share
June 7, 2006 12:37:34 AM

Real Conroe Benches.

Real benches, not Intel monitored.

More about : real conroe benches

June 7, 2006 12:44:05 AM

Quote:
Real Conroe Benches.

Real benches, not Intel monitored.


0Mg @Md p4wn5 4LL 4X4 wILL Roll 1NteL TO +H3 CURB j00 noObS. 4N@nd+3cH 15 In 1N+3L5 poCkE+ J00 N0O85 Kn0W nOTH1Ng, CONro3 w1LL $UXOr b3c@U$E 0F C@cH3 tHR4$h1N9 @ND WIlL N3vEr 3ven b3@T 4 3800 X2 NOOB 1DIots. D0 u5 4LL @ F4V0R 1N+El1Od5, amD aM2 4x4 @nD thE K8l w1ll p4Wn c0r03 4nD CoROE 2 @ND CONR03 3. B0w dOWN +O 4MD N00b$ mMM I$ Go0d 4ND G-inCH r0ck$!!!
Related resources
a c 478 à CPUs
a c 117 å Intel
June 7, 2006 12:54:23 AM

Finally, some real benchmarks!!

I'm sure some AMD fanboy will post that the FX-62 trouced the Conroe Extreme in the ScienceMark 2.0 - Memory Latency and Memory Bandwidth tests though and is therefore superior to Conroe. They'll probably say that Conroe isn't even worth the silicon used to make the chip.

Fanboys... can't live with them, can't kill them.
June 7, 2006 1:02:15 AM

Quote:
Finally, some real benchmarks!!

I'm sure some AMD fanboy will post that the FX-62 trouced the Conroe Extreme in the ScienceMark 2.0 - Memory Latency and Memory Bandwidth tests though and is therefore superior to Conroe. They'll probably say that Conroe isn't even worth the silicon used to make the chip.

Fanboys... can't live with them, can't kill them.


4X4 WILL P@WN J00 @nD J00R cuMsP0T 2, j00 N0O8s DOnT KN0W 4nYtH1N9 @mD WILL h4ve 0.65 oUt 8y JuLY 23 @nd @Md will 83 5p4nkIng J00 l1K3 I $PanKEd J00R $I5+eR HAR H4R H4R N0O8. in+3L pHAK35 b3nchM@rk5 4nd 5+3ALS SmALl CHIDr3n pHrOM +H31r h0M3S 4ND MaKe$ +H3m WORK 1N tHE pH@8s, iNKW1R3r $4Y$ $0O. mmM w45 R19HT J00 n00b$ DONt Know 4NyThiN9 4N4nD+3Ch 15 RUN MY c0MMuN15+S 4Nd C@NT b3 trU$TED J00 5TupiD n00b$. 4mD r0X0R3z J00 4lL 5MELl LIKE PH15h 5+uPid 1n+3LO1D5 Go B4ThE.
June 7, 2006 1:10:47 AM

:?: I see you got a face lift. IS THIS REALLY YOU?
June 7, 2006 1:13:55 AM

Quote:
Real Conroe Benches.

Real benches, not Intel monitored.


0Mg @Md p4wn5 4LL 4X4 wILL Roll 1NteL TO +H3 CURB j00 noObS. 4N@nd+3cH 15 In 1N+3L5 poCkE+ J00 N0O85 Kn0W nOTH1Ng, CONro3 w1LL $UXOr b3c@U$E 0F C@cH3 tHR4$h1N9 @ND WIlL N3vEr 3ven b3@T 4 3800 X2 NOOB 1DIots. D0 u5 4LL @ F4V0R 1N+El1Od5, amD aM2 4x4 @nD thE K8l w1ll p4Wn c0r03 4nD CoROE 2 @ND CONR03 3. B0w dOWN +O 4MD N00b$ mMM I$ Go0d 4ND G-inCH r0ck$!!!

Was all that really necessary? Surprisingly I can actually read what it says although most of the time I don't even bother trying. How long did that take? 8O

Edit: I finished reading the article. Interesting, and definately a step in the right direction for intel.
-mcg
June 7, 2006 1:18:32 AM

Quote:
Real Conroe Benches.

Real benches, not Intel monitored.


Indeed, it is starting. Here is a quick analysis of the data with some info for other flagship Intel processors thrown in for comparison:

June 7, 2006 1:30:49 AM

Well so be it. Don't see any reason to shit my pants. When it comes out I'm sure we'll all buy one.
June 7, 2006 1:32:32 AM

Quote:
Real Conroe Benches.

Real benches, not Intel monitored.
Any HORDE wish to contest? MMM? Anyone? Please, I do love to read your unsubstantiated, illogical AMD patronistic posts.
June 7, 2006 1:41:06 AM

Quote:
Finally, some real benchmarks!!

I'm sure some AMD fanboy will post that the FX-62 trouced the Conroe Extreme in the ScienceMark 2.0 - Memory Latency and Memory Bandwidth tests though and is therefore superior to Conroe. They'll probably say that Conroe isn't even worth the silicon used to make the chip.

Fanboys... can't live with them, can't kill them.


Lol, did you even look at the game benchies? AMD won, hands down and Tom even concluded FX was better. Now I'm not saying that Core 2 won't be good, just that K8L's will be better; if current 90nm K8s can beat 65nm Core 2s then K8L should beat Core 2, with four cores (or mabey two; LET US KNOW AMD!) and a shared L3 cache K8L's WILL have better preformance.

Its science, you can't argue with it

BTW: I'm not a fanboy, I just look into the future; I'm really happy that Intel is giving their best, but they need a much better architecture to beat AMD; I'm more a faN of AMD than of Intel because of one fact; Intel is this massive company and has put the processor market in a stranglehold
June 7, 2006 1:44:12 AM

I won't. Even though the performance is probably enough reason for me to upgrade my CPU, I'd have to change out my whole platform, and I have too much money invested in memory/mobo... and too little to shell out for new DDR2 and a decent motherboard. I'll probably sit this generation out.

-mcg
June 7, 2006 1:50:18 AM

Quote:


Lol, did you even look at the game benchies? AMD won, hands down and Tom even concluded FX was better. Now I'm not saying that Core 2 won't be good, just that K8L's will be better; if current 90nm K8s can beat 65nm Core 2s then K8L should beat Core 2, with four cores (or mabey two; LET US KNOW AMD!) and a shared L3 cache K8L's WILL have better preformance.

Its science, you can't argue with it

BTW: I'm not a fanboy, I just look into the future; I'm really happy that Intel is giving their best, but they need a much better architecture to beat AMD; I'm more a faN of AMD than of Intel because of one fact; Intel is this massive company and has put the processor market in a stranglehold


Are you on crack? where the the FX beat Conroe? did you even bother to look at the link above?

The Conroe won in every game bench and by large margin too..

I don't know if you were serious or just sarcastic. but DAMN!!!!
a c 478 à CPUs
a c 117 å Intel
June 7, 2006 1:50:50 AM

Quote:
Finally, some real benchmarks!!

I'm sure some AMD fanboy will post that the FX-62 trouced the Conroe Extreme in the ScienceMark 2.0 - Memory Latency and Memory Bandwidth tests though and is therefore superior to Conroe. They'll probably say that Conroe isn't even worth the silicon used to make the chip.

Fanboys... can't live with them, can't kill them.


Lol, did you even look at the game benchies? AMD won, hands down and Tom even concluded FX was better. Now I'm not saying that Core 2 won't be good, just that K8L's will be better; if current 90nm K8s can beat 65nm Core 2s then K8L should beat Core 2, with four cores (or mabey two; LET US KNOW AMD!) and a shared L3 cache K8L's WILL have better preformance.



Yes, I did look at the benchmarks.

You should feel real stupid though, because the link is to Anandtech's benchmarks.

I did read the review.

Apparently, you did not.
June 7, 2006 1:53:16 AM

Nice graphs !

My only criticism is in relation to the TDP values that you used.

While the TDP you used for the FX62 is correct, you need to keep in mind that AMD does rate the maximum theorical TDP of their CPUs while Intel (for their Netburst CPUs AFAIK) decrease that maximum theorical value by 25%.

On top of that, the TDP for the 965EE, 955EE and 840EE is 130W

Source

Quote:
Athlon 64 notes :

Thermal Design Power (TDP) is measured under the conditions of TcaseMax, IDDMax, and VDD=VID_VDD, and include all power dissipated on-die from VDD, VDDIO, VLDT, VTT, and VDDA.


Quote:
P4/P4 EE Notes:

The Estimated Max Power Dissipation numbers are based on the fact that Intel estimates the power dissipation for various software applications and sets the Thermal Design Point as the upper limit for how far these applications might push the Pentium 4. Intel states: "Processor power dissipation simulations indicate a maximum application power in the range of 75% of the maximum power for a given frequency." So the Est Max Power Diss was calculated by dividing the TDP by 75%


Which leads me to believe that either the FX62 TDP should be adjusted to 94W (125W x 0.75) or that the three EE P4's should be adjusted to 173W (130W x 1.33) in order to make a fair comparaison.
June 7, 2006 2:03:04 AM

smokinrsx ARE U ON CRACK? HE POSTED THAT AFTER jaguarskx DID. COMMON SENSE!
June 7, 2006 2:09:38 AM

Quote:
Finally, some real benchmarks!!

I'm sure some AMD fanboy will post that the FX-62 trouced the Conroe Extreme in the ScienceMark 2.0 - Memory Latency and Memory Bandwidth tests though and is therefore superior to Conroe. They'll probably say that Conroe isn't even worth the silicon used to make the chip.

Fanboys... can't live with them, can't kill them.


Lol, did you even look at the game benchies? AMD won, hands down and Tom even concluded FX was better. Now I'm not saying that Core 2 won't be good, just that K8L's will be better; if current 90nm K8s can beat 65nm Core 2s then K8L should beat Core 2, with four cores (or mabey two; LET US KNOW AMD!) and a shared L3 cache K8L's WILL have better preformance.

Its science, you can't argue with it

BTW: I'm not a fanboy, I just look into the future; I'm really happy that Intel is giving their best, but they need a much better architecture to beat AMD; I'm more a faN of AMD than of Intel because of one fact; Intel is this massive company and has put the processor market in a stranglehold

Science typically dictates that when the numbers say oen thing is better, it typically is.

http://anandtech.com/cpuchipsets/showdoc.aspx?i=2771&p=...

I think you need to look at that page again because you obviously missed out on the Conroe 0wn1ng the FX-62 in every bench.

From TOM'S HARDWARE:
Quote:
. . . we were able to determine that not even AMD's top-of-the-line Athlon FX-62 CPU running at 3.0 GHz could clearly best the pre-release model of the Core 2 Duo (2.66 GHz Conroe) processor we tested. This comparison didn't even use the top-of-the-line Conroe processor, which Intel plans to introduce soon.


You are as blind and loyal as 9 nm.

And OOOOOO YEAH L3 CACHE WOOOO. L3 cache is trash. It has worse latency than 14.4K dialup. Anybody who swaps L2 for L3 is a tard. Seriously, what a bad idea. The K8 core lost every benchmark except the memory tests (and yeah, science mark is basically that).

AMD is the one playing catch up here, and it is unlikely that the K8L will be anything special, as it is based off the same damn architecture. They had a good design. Intel's is better. They have to work on something new.
June 7, 2006 2:15:26 AM

Quote:
Which leads me to believe that either the FX62 TDP should be adjusted to 94W (125W x 0.75) or that the three EE P4's should be adjusted to 173W (130W x 1.33) in order to make a fair comparaison.


No, not really that different. Look at AM2 FX-62 benches and you'll see latest AM2 FX's and latest Pentium EE's are close to each other in power consumption.
June 7, 2006 2:16:47 AM

Quote:
Finally, some real benchmarks!!

I'm sure some AMD fanboy will post that the FX-62 trouced the Conroe Extreme in the ScienceMark 2.0 - Memory Latency and Memory Bandwidth tests though and is therefore superior to Conroe. They'll probably say that Conroe isn't even worth the silicon used to make the chip.

Fanboys... can't live with them, can't kill them.


Yeah, but it is pretty damn easy to make them look completely, utterly, totally, and undeniably stupid. I blame in on the parents and the education system. I blame it on the Canadians :tongue: .
June 7, 2006 2:21:18 AM

I said Tomshardware's comparions are flawed because they took IDF system's results and compare with their lab results. And I was right. Which is one big mistake they can make. They often do that. Most of the times they don't but Tomshardware does make stupid comparisons. See how Anandtech's and Hexus's comparisons differ so much from Tom's.
June 7, 2006 2:27:05 AM

Quote:
Real Conroe Benches.

Real benches, not Intel monitored.
Any HORDE wish to contest? MMM? Anyone? Please, I do love to read your unsubstantiated, illogical AMD patronistic posts.

S7A88Y is certainly looking for membership wouldn't you say :)  ... dang how can he say these things and not feel completely embarrased :oops:  I suppose, but I am certainly not looking to rub it in. He has suffered quite enough.
June 7, 2006 2:30:57 AM

Quote:
Which leads me to believe that either the FX62 TDP should be adjusted to 94W (125W x 0.75) or that the three EE P4's should be adjusted to 173W (130W x 1.33) in order to make a fair comparaison.


NOT FOR CORE. If you actually care to go around and see Pentium M comparisons a 27W Pentium M is only 25W in ACTUAL POWER USAGE!!!

The reason there is such a difference between Pentium M and Pentium D/4 is BECAUSE OF POWER MANAGEMENT.

EIST on Pentium M is SUPERIOR to EIST on Pentium 4/D. Core 2 has EVEN BETTER power management than Pentium M.

Well, you took the care to quote me, where did I mention Core and Pentium M ?

Right, I did'nt...

So, no need to BOLD OUT AND CAPITALISE what you think is IMPORTANT in your POST, most readers here are MORE THAN CAPABLE of figuring that out BY THEMSELVES.

Annoying eeh ?

*Edit : Good thing that I quoted him for posterity :

Last edited by DavidC1 on Tue Jun 06, 2006 10:17 pm; edited 1 time in total
June 7, 2006 2:43:04 AM

According to xbit its 135w for the FX 62.

Link.
June 7, 2006 3:05:40 AM

Quote:
According to xbit its 135w for the FX 62.

Link.


Very interesting, 156W for the B1 955EE...

Quote:
We measured the current that goes through the CPU power circuitry. So, the numbers given below do not take into account the efficiency of the CPU voltage regulator laid out on the mainboard.


All we need to know is that the efficiency of the VRMs is for the boards they used.
June 7, 2006 3:17:30 AM

Updated the data to reflect the typo I had in the FX-62 price, and the slightly higher TDPs for the Pentiums. I'm going with the manufacturer reported TDP because data can always be skewed either way.

Doesn't really change how the graph looks, but thanks for the suggestions.

June 7, 2006 3:45:31 AM

Ooh, no big deal really, it is clear enough that the Core 2 uarch is more than competitive no matter how you look at it, I'm glad that Netburst's days are numbered, that is, if Intel can manage to supply the insane demand there will be for their new products while keeping the prices resoneable.
June 7, 2006 3:45:48 AM

Wow! Very, very impressive!!!!

I am having some trouble interpreting some of this new data, though. I was wondering if itguy or 9-inch could comment on this new data so that I can be sure I fully understand it. Please help me here 9-inch or itguy. These benchmarks just can't be correct. The FX 62 was supposed to win hands down due to the conroe's cash trashing. Please inform me on this very urgent matter.
June 7, 2006 4:31:34 AM

Those are good results to say the least. What I found most interesting though was how close the X6800 is to the FX62 in memory latency. I guess there is still some mileage left in the FSB. I bet with a 1333MHz FSB, Conroe's memory latency may actually be lower than the FX's. Woodcrest wouldn't show the same results though because of the FB-DIMMs. I think a lot of this can be attributed to the redesigned memory controller in the 965 chipsets and Fast Memory Access, which just means that i975 results will probably be disappointing.

The memory bandwidth results are disappointing, but those should improve when the final revisions and later BIOS updates are released. The current P965 chipsets are still engineering samples.

http://www.theinquirer.net/?article=32227

It seems Conroe itself is ready to launch, which should mean volumes and yields are great, but is being held back by the availability of the mainstream 965 chipsets. Production chipsets just aren't available yet. This probably isn't a design issue, but just a consequence of the transition to 90nm 300mm wafers from 130nm and 200mm wafers. It'll take them a little bit to figure out the best way to go about it and get the yields up. Ironically, this is precisely the reason why Intel's policy was to avoid trying to transition new logic and a new process at the same time since issues are compounded. Sadly, they've broken that policy with Yonah, which was why availability was initially low, and they've done it again with the 965 chipsets.

As they revise the process, hopefully they'll fix the power consumption issue of the new 965 chipset.

http://www.theinquirer.net/?article=32230

Transitioning from 130nm to 90nm should have reduced both power consumption and temperatures. I think it's possible that the early engineering samples are built on the 130nm process, which would explain why the power consumption is so high. It's wierd that the power levels of the i975 is so low, below even the i945P.
a c 100 à CPUs
June 7, 2006 4:46:13 AM

Not to nitpick, but your TDP is normalized with the 840EE being 1.08, not 1.00. You probably need to divide every TDP number by 130 now instead. No biggie.
June 7, 2006 5:18:15 AM

Quote:
Finally, some real benchmarks!!

I'm sure some AMD fanboy will post that the FX-62 trouced the Conroe Extreme in the ScienceMark 2.0 - Memory Latency and Memory Bandwidth tests though and is therefore superior to Conroe. They'll probably say that Conroe isn't even worth the silicon used to make the chip.

Fanboys... can't live with them, can't kill them.


4X4 WILL P@WN J00 @nD J00R cuMsP0T 2, j00 N0O8s DOnT KN0W 4nYtH1N9 @mD WILL h4ve 0.65 oUt 8y JuLY 23 @nd @Md will 83 5p4nkIng J00 l1K3 I $PanKEd J00R $I5+eR HAR H4R H4R N0O8. in+3L pHAK35 b3nchM@rk5 4nd 5+3ALS SmALl CHIDr3n pHrOM +H31r h0M3S 4ND MaKe$ +H3m WORK 1N tHE pH@8s, iNKW1R3r $4Y$ $0O. mmM w45 R19HT J00 n00b$ DONt Know 4NyThiN9 4N4nD+3Ch 15 RUN MY c0MMuN15+S 4Nd C@NT b3 trU$TED J00 5TupiD n00b$. 4mD r0X0R3z J00 4lL 5MELl LIKE PH15h 5+uPid 1n+3LO1D5 Go B4ThE.

excuse me, what was that you said about me and my fellow comrades?
June 7, 2006 5:29:38 AM

Finally some semi independent benchmarks that show Conroe being impressive.

However, what stuck out at me was the rollout plan - only 25% of CPUs shipped in the 4th quarter appear to be core 2 duo based (I would bet this is more due to overstocks on Netburst chips than Conroe shortages). However, Intel will have to balance this strategy carefully, as they need to equalize or regain their market share as soon as they can to improve their financials.

As the article said, this may also serve to drive up prices in the consumer market as well, which is bad for early adopters.
June 7, 2006 5:30:47 AM

I suppose I am Rich! So I will settle for the E6600... when it drops to $299! :wink:
June 7, 2006 5:38:08 AM

Quote:
I suppose I am Rich! So I will settle for the E6600... when it drops to $299! :wink:
That is only a $17 difference.
June 7, 2006 6:21:58 AM

That is some nice bechies with the Core 2 EE and the rest of the family and what's the great thing also is that the P4, 800, and 900 chips are just keep getting cheaper every month. This is great! :D 
June 7, 2006 7:27:06 AM

Quote:
That is some nice bechies with the Core 2 EE and the rest of the family and what's the great thing also is that the P4, 800, and 900 chips are just keep getting cheaper every month. This is great! :D 


Well, the logic follows, if the independent benches are showing the same performance gains as IDF, and the "intel" controlled benches, then we can assume that the "intel" controlled benches were in fact not "rigged".... Hmmm I wonder if 9-inch would accept that logic :) 

Intel has honestly given up on the Netburst line and want to get it off the books ASAP -- I would not be surprised if the Netburst loading in all the fabs are 0 right now and 100% is either Yohna, Conroe, Merom or Woodcrest. It will be interesting to see if Intel meets or beats it's shipment percentages -- if not already done, they should simply stop all Cedarmill, Presler, Prescott, Smithfield production and start all new arch.

Jack

There is actually an economic argument to keep some Netburst product around.

The third world and pacific rim are still big consumers of low end chips, well under $100 typically, and you don't want to devalue Conroe by having it play in that market.

The cost of producing a 65 nano PentiumD is likely pretty similar to that of a Conroe, less if you factor in that the PentiumD kit is all paid for already.

Also, may upgraders simply can't/won't step up to the pump twice to get a new motherboard and chip.

Intel seems to be starting to really firesale the 9xx series, and frankly PentiumD 940 (2 x 3.2 ghz) at $183 will likely outrun a 1.86 ghz Conroe at the same price....

The 975x and 965 boards are likely to be pretty pricey for a while, but stuffing a bunch of 940s into all the 945 boards kicking around seems like good bang for the buck to me.

Wonder what AMD will do with their x2 series.... FX60s for $299 anybody?

~~~~~

A FYI regarding the AnandTEch benches - Other than the two symthetic memory tests, Conroe won all 14 benchmarks by a low of 3% to a high of 52% - The average margin of victory was 24%.

BY comparison, the FX60, on average, only beats the EE965 by 7% on a similar set of benches.

I believe this is the largest performance delta between Intel and AMD since the old AMD T-Bird was slapping around the early P4 Willamete kludges.
June 7, 2006 7:40:43 AM

I can't believe you still dream of a 173w TDP for Pentium D.
If Intel said that the maximum amount of working transistors in their CPU at one time is 75%, then it is 75% because they know their cpus. You can't take AMD design and compare it to Intel's because everyone knows that a Pentium D is a totally different animal from a 64 X2. That means if 100% of X2's transistors are working @ 100% load doesn't mean that the D will do the same. May be it is impossible to get 100% of D's transistor to work all at once.

TDP rating is used both by Intel and AMD to let users know how much heat they need to dissipate when using certain CPUs. Who would be stupid enough to use a 130 TDP rating HSF on a 173 w CPU, the one you dream up?

TDP is not equal to current * voltage. Samething like the heat given off from a lamp is not equal to current * voltage going into it. Some of the power is converted to light.
June 7, 2006 7:42:50 AM

It can only get better, since competition from Intel means that AMD will make better processors, and best of all the prices drop.
a c 100 à CPUs
June 7, 2006 1:06:30 PM

No, I think that it has to do with the history of the two companies and also will explain a few differences in running a NetBurst chip and a K8.

AMD used to have some pretty bad heat problems back in the day (IIRC some of the K6 line and one of the 180nm K7 variants had big trouble with running very hot) while Intel never really did. So AMD, wanting to shed its image of being unreliable, rated its chips at an absolute worst-case scenario usage so that manufacturers would be sure to put enough cooling on the chips to ensure that they ran at a decent temp. Intel didn't really have any especially hot-running chips before the NetBurst line, and it used its lower 75% UL standard TDP rating method because most people do not run a chip at 100% load 24/7 and even if they did, their chips didn't run that hot and a handful more watts could be easily dissipated.

Intel's Northwood P4C at 3.0GHz ran pretty equal to a tad better than the Athlon XP 3200+ and the 3.2C and 3.4C were simply faster than any AMD chip at the time. The Northwood line started to get a little warm near the end, but it was not that big of a problem as they were the performance leaders. And a die shrink to 90nm was coming, so Intel thought that the clock speed would keep going up and the temps would go down. Now enter the K8. The fastest K8s beat the 3.4C and Intel's response was to make the Prescott. However, the Prescott actually ran hotter than the Northwoods at the same clockspeed. With the K8s beating the P4s in performance and TDP, Intel sure as heck wasn't going to up-rate the TDP on its chips as that would be marketing and sales suicide. It would also cast doubts as to how high the Prescott would scale, and Intel promised >4GHz, which would keep them somewhat competitive with AMD. The Prescott ended up only gaining 400 MHz over the Northwoods and ran into throttling problems when run under load, but one could overclock a Prescott to a decent margin or get them to run as cool as the K8s if they put a much larger cooler on the CPU.

It was basically Intel's stubbornness to admit defeat to AMD and an attempt to save face with an architecture that had flaws that led Intel to keep the TDP lower than it should have been for the P4 line. Now Intel doesn't want to redo their method of computing TDP with any other of their chips as they made performance per watt a very played-up marketing tool since their Pentium M/Core/Core 2 line runs with a rather low TDP. I think that Intel will never revise their TDP methods as it would simply give ammunition to AMD. I also think that AMD won't move from their 100% load TDP to the 75% UL standard that Intel does so that its chips will be reliably cooled and be more durable, not to mention run cooler in operation. AMD's main advertisement method is word of mouth and reviews, so their chips must run better while in operation than Intel's for that advertisement method to work: i.e. a 50W TDP Intel desktop chip should will run hotter than a 50W TDP AMD chip and that will get noticed.
June 7, 2006 1:44:05 PM

Anandtech used CL5 memory. That may be why memory latency is so close. Lower latency would help the FX-62 much more than the X6800. They should have used CL4, it's the norm for DDR2-800.
June 7, 2006 1:49:48 PM

Since you spent sometime to reply to my post, I decided to dig around the net to pull up some old numbers:

http://techreport.com/reviews/2004q1/p4-prescott/index.x?pg=2

Quote:
Northwood's TDP at 3.2GHz is 82W, while the Extreme Edition's is about 92W. Prescott's TDP at 3.2GHz is 103W. So yeah, this thing pulls some juice and generates some heat.

To manage Prescott's thermal prowess, Intel has created a new specification for thermals that allows for finer-grained control of fan speeds based on a value returned from the CPU. This value is set "based on the power dissipation of each unit," according to Intel, and combined with the thermal diode temp, will dictate safe fan speeds for coolers. Implementing this scheme will require motherboard changes, but not changes to the actual cooler designs. In fact, Intel-approved coolers for current Pentium 4s should work for Prescott at its initial speed grades.


And:

http://www.xbitlabs.com/news/cpu/display/20050212034615.html

Quote:
Intel Pentium 4 processors 630, 640 and 650 are compatible with mainboards sporting Intel’s 04A Platform Compatibility Guide (PGC), e.g., they have Thermal Design Power (TDP) of 84W; Top-of-the-range Intel Pentium 4 processor model 660 complies to 04B PGC, which means that the chip may have TDP of up to 115W, as previously expected.


And:

http://www.digital-daily.com/cpu/intel_pentium_d_9x0/

D820 : 95 W
D830 & D840: 130 W
D920&930 B1: 95 W
D940&950 B1: 130 W
D930&940&950 C1: 95 W

It seems to me that the TDP numbers given from Intel are logically increasing from Northwood to Prescott to Presler. You must remember that TDP rating is not a number to be joke around or lie about. It is not a performance number for users to look at. It is a very important number, which most end users don't even know or ask about, that many vendors who use or build computer around the CPU need to know so they can do a proper heat dissipation job. I doubt any chip maker would dare to lie about the TDP rating unless they want their chips to go up in smoke in a consumer product.

Quote:
Intel didn't really have any especially hot-running chips before the NetBurst line, and it used its lower 75% UL standard TDP rating method because most people do not run a chip at 100% load 24/7 and even if they did, their chips didn't run that hot and a handful more watts could be easily dissipated.


True, but you must remember, it doesn't take that long for a little extra heat to build up into a real copper melting temperature. If 173 w was really the TDP of the D running at full speed, using a 130 w rating HFS will kill it in seconds with that extra 40 w of heat. Ever use a 30 W soldering iron? you will know what I mean.

Like I said, AMD designed their cpu totally different from Intel's. AMDs' having 100% transistors working under 100% load doesn't mean Intels' do too.
June 7, 2006 2:10:01 PM

~0.1 Ghz clock difference but a whooping >20% performance difference, great results. I have to admit though that with low lcs the difference might shrink by a few percents but that's about it.
Also it's really funny to see all those AMD n00bs saying that they wanted tests at high resolutions were the "FSB would ruin performance". Yes the difference at high res. is diminished,but not because of CPU power but because of GPU bottleneck. Such fanboyism.
June 7, 2006 2:10:42 PM

father.. please forgive me.. for i've sinned..<------*ok... wtf?*
man.. seeing is really believing
i was an AMD fanboy, probably a hardcore one. but after what i've saw what Conroe can do, i have to say i'm AMD fanboy no more!!

i attended the computex taipei *i live in taiwan*, and actually get to play with the both Conroe and FX-62 systems a little bit, while intel representatives weren't watching. Conroe basically OWNS FX-62, in almost every single benchmark, game, and application. since Computex does not allow non-media personnel to take pictures, i'm sorry to say i can't share the amazement with you guys.

something tomshardware and inquirer had not mentioned. Conroe actually can run withOUT CPU fan, under 35% load, for a significant amount of time. Conroe was automatically throttled to 1.8Ghz. of course, doing that would potentially damage the processor chip, but Conroe actually can perform without CPU fan.

i remembered a test done by HKEPC, where they shut down the CPU fan on a X2 3800+. the CPU worked for 7 min before it crashes. however, on a Conroe system, it still runs smoothly over an hour.

once i heard someone talking about how Conroe is the perfect x86 CPU. i didn't believe back then. today, i'm starting to believe.
June 7, 2006 2:20:42 PM

Quote:
Finally some semi independent benchmarks that show Conroe being impressive.

However, what stuck out at me was the rollout plan - only 25% of CPUs shipped in the 4th quarter appear to be core 2 duo based (I would bet this is more due to overstocks on Netburst chips than Conroe shortages). However, Intel will have to balance this strategy carefully, as they need to equalize or regain their market share as soon as they can to improve their financials.

As the article said, this may also serve to drive up prices in the consumer market as well, which is bad for early adopters.


Well, if you are beginning to accept the semi-independent (actually independent -- Intel was NOT involved) benches as being impressive then looking at the numbers it seems to be validating all the other data that most otherwise claimed was invalid does it not?

If the independent benches are validating that data, then we can stretch the assumption that the hyper-critical-non-real-performance-benched-because-it-was-intel point of veiw was not warranted hence there is actually a huge body of valid data showing Core 2 superiority to date. All the IDF benches were in fact not rigged, the Xtremesystem OCers were right that this is a fast chip, and the Hexus nuts did in fact do a fair compare.

When all signs pointing in a direction say east, you look to the east in the morning and see the sun, then it is a good chance that the signs were right :)  .....

Jack

When did I say I didn't believe the performance data? I agreed that the numbers were impressive all along. I just said I took Intel's benchmarks with a grain of salt and wanted to see independent ones. The only reason I called these semi-independent is because they were not able to do the full body of tests that they would normally do. It is interesting (and expected), that the gains shown in these benches are a little less drastic than those originally shown by Intel.

I don't particularly care who wins, I just like to watch. My PC is recent enough that I likely won't be upgrading for 18-24 anyway.
June 7, 2006 5:40:17 PM

Quote:
Finally, some real benchmarks!!

I'm sure some AMD fanboy will post that the FX-62 trouced the Conroe Extreme in the ScienceMark 2.0 - Memory Latency and Memory Bandwidth tests though and is therefore superior to Conroe. They'll probably say that Conroe isn't even worth the silicon used to make the chip.

Fanboys... can't live with them, can't kill them.



Actually, that was a real win. The only thing about the ScienceMark bandwidth is that it means AMD has A LOT of room to improve efficiency. If they can get some IPC out that bandwidth, it will be serious.
June 7, 2006 5:58:30 PM

Quote:
father.. please forgive me.. for i've sinned..<------*ok... wtf?*
man.. seeing is really believing
i was an AMD fanboy, probably a hardcore one. but after what i've saw what Conroe can do, i have to say i'm AMD fanboy no more!!

i attended the computex taipei *i live in taiwan*, and actually get to play with the both Conroe and FX-62 systems a little bit, while intel representatives weren't watching. Conroe basically OWNS FX-62, in almost every single benchmark, game, and application. since Computex does not allow non-media personnel to take pictures, i'm sorry to say i can't share the amazement with you guys.

something tomshardware and inquirer had not mentioned. Conroe actually can run withOUT CPU fan, under 35% load, for a significant amount of time. Conroe was automatically throttled to 1.8Ghz. of course, doing that would potentially damage the processor chip, but Conroe actually can perform without CPU fan.

i remembered a test done by HKEPC, where they shut down the CPU fan on a X2 3800+. the CPU worked for 7 min before it crashes. however, on a Conroe system, it still runs smoothly over an hour.

once i heard someone talking about how Conroe is the perfect x86 CPU. i didn't believe back then. today, i'm starting to believe.
Thats the way it works in the CPU industry, in mabey two to three years from now, AMD will (hopefully) have something that kicks Conroe in the balls. Leapfrog :)  .
June 7, 2006 5:58:45 PM

anyone can tell me if im the only one that noticed something fishy on Anand's benchmarks?
why you must ask?
well because like 4-6 benchmarks sites I've seen wthat only tested the Fx-62 had a Winstone scores of 32-33 and similar scores
why anand's is always LOWER than most of other benchmarks??
June 7, 2006 6:05:16 PM

Quote:
Finally, some real benchmarks!!

I'm sure some AMD fanboy will post that the FX-62 trouced the Conroe Extreme in the ScienceMark 2.0 - Memory Latency and Memory Bandwidth tests though and is therefore superior to Conroe. They'll probably say that Conroe isn't even worth the silicon used to make the chip.

Fanboys... can't live with them, can't kill them.



Actually, that was a real win. The only thing about the ScienceMark bandwidth is that it means AMD has A LOT of room to improve efficiency. If they can get some IPC out that bandwidth, it will be serious.

Or it's a sign that Core 2 is more memory efficient than the K8 is.
June 7, 2006 6:09:54 PM

Quote:
anyone can tell me if im the only one that noticed something fishy on Anand's benchmarks?
why you must ask?
well because like 4-6 benchmarks sites I've seen wthat only tested the Fx-62 had a Winstone scores of 32-33 and similar scores
why anand's is always LOWER than most of other benchmarks??
Yup, your right. All credible sites have suddenly jumped on the Intel bandwagon, and are out to get AMD with every bit of incorrect, Pro-Intel information as possible. Your right, everyone is out to get AMD with these benchmarks.
!