HEXUS.reviews :: WORLD EXCLUSIVE! INTEL CONROE BENCHMARKED

[Bob Crabtree]

David,

The percentage relative calculations for those results expressed in units of time ("lower is better") are wrong in the original article, for the reason I gave. (I addressed my explanation to Bob, inasmuch as the author of the original article did not comment on my post, whereas Bob did.)

David,

Please don't shoot the messenger (ie mxyztplk).

mxyztplk's posting of 06:54 PM today does hit a number of nails rather firmly on their topmost flat area.

I'm looking into this myself because, subsequent to my substantive posting (and before reading what you or mxyztplk had subsequently said) I'd come to believe that we probably have not presented this percentage info in the best possible way for clarity of reading - though that's not something any of us should beat ourselves up over (or allow anyone else to do) given the hugely compressed timing of the builds, the tests, the write ups and the publication of them all.

In addition, I have now come to the same conclusion as mxyztplk about the figures which involve time (contrary to my earlier, rather off-hand dismissal of mxyztplk's initial comments. Sorry mxyztplk!).

I need to communicate with the author, Mr Deeplung, so that we can sort this out once and for all - because we do need to ensure that we correct any errors and make quite clear the meaning of the information that we include.

Truth is, though, my brain has now ceased up (doubtless Mr D's will be in the same state or worse), so I'll try to ensure that we talk tomorrow.

For everyone else reading this - apologies for any confusion (and, again, for my earlier off-hand dismissal of mxyztplk's initial posting).

The best thing to do in the interim - ie until we sort out a better and more accurate presentation of the differentials - is simply to disregard the worked up figures and look at the raw figures in the table above. They, I think, are not in any way likely to mislead you.

Bob

[moritori]

The intel chip is indeed fast but there is a weakness that was brought to my attention from a Electrical Engineering Geek Guru. From there observation of the structure and how the CPU accesses the L2 cache memory it is his belief that they have tuned the chip to the operating system!!! What you say?
That is right, built a chip to work with a specific operating system! It is a great idea and one that deserves applause. Well done intel!

Unfortunatly that means when Vista comes out it will not "perform" or even work properly.
(were sorry but you will need to upgrade to our new chip... to use this operating system)

What a money maker! Wow can you image how much more revenue you will generate with this in place? Every 2 to 3 years you HAVE TO UPGRADE your system to use the next generation Operating system.

Haven't any of you wondered why they kept delaying the new Operating System, Vista?
Yeah Yeah it is because of codeing problems blah blah blah, but when it comes down to it it has never stopped Microsoft in the past.... ever heard of patching? hehe

It would be of great to see how the new intel chips perform with Windows 98 or even linux versus and AMD chip on those operating system.
http://forums.hexus.net/images/icons/icon10.gif
Talking

[DR]

Guess we got to test that...

[Bob Crabtree]

Just as a holding operation, these are the corrected percentage figures for the second table and relate only to results that were shown in the top table in units of time.

Corrected**********************************Original
FX-62/965*********************************FX-62/965
-2.12%/54.53% ** ScienceMark 2.0 memory latency * 2.16%/35.29%
26.41%/62.35% ** HEXUS Pifast****************20.89%/38.40%
31.20%/22.40% ** DivX encode - multithreaded*****23.78%/18.30%
32.46%/30.70% ** WAV conversion multithreaded***24.5%/23.49%

As you can see - and commiserations to AMD - the performance of Conroe E6700 was actually even more stellar than our original percentage figures suggested!

Willy his very self will try to sort out the original article tomorrow so that it carries these corrected figures.

Our thanks to mxyztplk for pointing out the problem (and for taking my initial dismissal of it withour rancour).

Bob

[mxyztplk]

Bob,

Thanks greatly for helping to get the relative percentage calculations sorted out.

On a different point, regarding the tests themselves: as has already been noted, it certainly would have been good to see a comparison using the same CPU DDR2 memory configuration, wherever possible. (E.g., in the test system configurations, the FX-62 memory configuration had 4 DIMMs vs Conroe's 2 DIMMs, 2GB vs 1GB, and 800MHz vs 667MHz.)

Among other effects, this major test system configuration advantage in favor of FX-62 obviously would/could have very greatly affected the results for those tests that are most heavily CPU memory-bandwidth dependent/constrained.

[Tarinder]

Thanks for the helping hand, mxyztplk.


Word on the internal grapevine is that Willy no longer has access to the Conroe system. It appears to have been pulled. We only had the CPUs for a short while and had to test them with the resources at our disposal during that timeframe, hence the slightly differing test-system configurations.

The purpose of the article, I believe, was to give our readers a quick snapshot of the general performance benefits of the upcoming Core 2 Duo when compared to the current FX-62 and E.E 965 processors.

You can be rest assured that, as AMD has moved on to utlising DDR2 for its AM2 form-factor, our full, in-depth review of Conroe's architecture and performance will have competing systems configured with identical components running at identical settings. The only difference will be the motherboard(s) and chipset drivers used.

[Bob Crabtree]

Bob,

Thanks greatly for helping to get the relative percentage calculations sorted out.

I'm pleased I had some input here and (I'm happy to say again) very pleased with your response to my earlier stupidity.

But Ol' Willy it was who went and sorted out the original article to get the figures correct - I've been watching the stuff on BBC 4 about climate change (eek!!)

I'm just amazed that Will had the energy after his recent testing and writing exertions!

On a different point, regarding the tests themselves: as has already been noted, it certainly would have been good to see a comparison using the same CPU DDR2 memory configuration, wherever possible. (E.g., in the test system configurations, the FX-62 memory configuration had 4 DIMMs vs Conroe's 2 DIMMs, 2GB vs 1GB, and 800MHz vs 667MHz.)

Among other effects, this major test system configuration advantage in favor of FX-62 obviously would/could have very greatly affected the results for those tests that are most heavily CPU memory-bandwidth dependent/constrained.

I won't bang on much here (and not simply because I don't understand the niceties of which you talk!), I'll just paraphrase what Tarinder said in answer to this and other queries of its ilk.

We did not do our usual fulsome and carefully balanced testing but simply the best we could with the kit we had available during the very short period in which all the CPUs were available to us - and, personally, I'm astonished just how well Ol' Willy acquitted himself.

And, as Tarinder has also said, we'll be revisiting Conroe vs AM2 in a lot more detail.

I would add that, as much as anything, what we were also trying to do was give people a clue about whether or not Intel's widely reported Conroe claims from the Spring 2006 IDF were reasonably valid and not totally cooked up.

I'd very much like to think that all the effort that's been involved has meant we've done just that (even though I'm sure I'm not the only AMD die-hard surprised to find out quite how amazing Conroe seems to be).

Bob

[Bob Crabtree]

another way to calc % chgs is [(A/B)-1]*100 which is mathematically identical to [(A-B)/B]*100. So anyone uses the 1st, bob's math is correct in that sense...
***on which is which
if you are looking to see the % differance between two objects A and B where you want to say that A is x% differant to B, then the maths is above.
If you say The percentages of how" Conroe performed relative to FX-62 ". then the maths is (C - FX / FX) of course, if the article reads "how FX performed compared to/relative to Conroe", then you should do the maths: (FX - C/C). I haven't dble-checked which way the article calls it, but hope this is clear.

btw. - why use =SUM in your formula?

merlin (hope you don't mind me shortening your name!)

Last thing first - and you are not the first person to ask that question (the last one was someone from BBC R&D who clearly also wasn't impressed!) - it's simply the result of the way that I use Excel.

When I was an Excel nubie, and wanted to do a calculation in Excel, I found that the easiest way was to start off by clicking on the AutoSum toolbar icon - Σ - and that, of course, automatically puts in the SUM after the = sign.

Why still work that way?

Well, even though I am a keyboard shortcut fanatic, what I realised is that it is still quicker to do what I do, rather than hit the = key and then move my hand from the keyboard to the mouse to select what cells were being acted upon.

And that's my excuse and I'm sticking to it!



As for your explanations on calculation methods - thanks for the input.

What you will have seen, though, is that some changes have been made as a result of comments - so, now, hopefully, things are as they should be.

Bob

[keta]

moritori , if they have done that then that wud b very stupid of them to do it so close to windows vista release? , now not everyone upgrades to a new up ? i been on windows xp since it came out and i dont plan on moving to vista till mid 2007 at the earliest , once all the bugs r fixd and aall the programs run fine then ill spend the money and move over , but if they have done the op thing , maybe they have programd it to work perfectly with vista and it just happens to work good on xp as well ? soo wat ur sayin cud b true , but it also cud give intel a even more performance increase if they have created this processor for vista , that wud mean much much more performance and i think if they was making a cpu for a op system then they wud make it for vista cuse its a vista ready cpu , it was ment to b lunchd with vista , but vista got delayd and am2 got moved forward soo they moved there release forward and now its coming out soon , soo im 90 percent sure that this processor wud have better performance on vista , specialy if wat u says is true. soo it wudn't be worse , it wud b betta

[BladeRnR]

I haven't upgraded for over a year now (Which is staggering when I think about it). I'm still running a Northwood 3.0Ghz on an 875 Chipset. Still solid and strong. I've been waiting (As we all do on occasion) for PC hardware to really surge ahead.

The other day I was in the MAC store of all places and was having a play with an iMac. Now I am OS / hardware agnostic (Being a coder & gamer) and was very excited to see this system able to boot into both WindowsXP and Mac OSX Tiger. Both were running natively of course.

Given there's a lot to like about both VISTA and MAX OSX (And with Leopard just around the corner) I've decided to purchase a PowerPC that may very well feature Conroe. Given Intel have also been given the task of creating the Motherboard for these systems one would hope they'd cater for the graphics and gamer crowd and bring out a 975X SLi mobo as well

Anyone care to comment on what the PowerPC specs might be? Imagine a system capable of booting into both Mac OSX and VISTA on an uber fast platform

More can be found here:

http://www.fortwayne.com/mld/newssen...g/14657483.htm

[JumpingJack]

Just as a holding operation, these are the corrected percentage figures for the second table and relate only to results that were shown in the top table in units of time.

Corrected**********************************Original
FX-62/965*********************************FX-62/965
-2.12%/54.53% ** ScienceMark 2.0 memory latency * 2.16%/35.29%
26.41%/62.35% ** HEXUS Pifast****************20.89%/38.40%
31.20%/22.40% ** DivX encode - multithreaded*****23.78%/18.30%
32.46%/30.70% ** WAV conversion multithreaded***24.5%/23.49%

As you can see - and commiserations to AMD - the performance of Conroe E6700 was actually even more stellar than our original percentage figures suggested!

Willy his very self will try to sort out the original article tomorrow so that it carries these corrected figures.

Our thanks to mxyztplk for pointing out the problem (and for taking my initial dismissal of it withour rancour).


Bob

The original figures are actually correct.... you are miscalculating based on the incorrect rule: always divide to the lower. This is not right.

If you are comparing against the FX-62 to say "how much faster will this DivX encode, the correct formulation is (FX62 - Core2)/FX62 * 100 so that comes to (164-125)/164 *100 = 23.8 %. (rounded to nearest 10th)

To check this correctly, if I took 23.8% of 164 (the FX62 rendering time) I should get the amount of time less that Core 2 will render. Thus the rendering time would be 164 - Core advantage.

Or

164 - 164(0.238) = 125.

This is a common mistake. You divide by what you are referencing to, in this case Core 2 will be X% faster than FX62 thus no matter the metric always normalize to the comparator.

Now even 23.8% is massively impressive, Core 2 is gonna a heck of a CPU.

Jack

[acrobat]

Anyone able to answer my questions please? They seem to have got lost a few pages back now. Here they are again:

1) Is the Conroe chip 64bit? --- If not, maybe AMD chips will have an advantage once Vista becomes widespread?

2) When is conroe due? Your article said, "It's abundantly clear that Conroe will be the fastest CPU in town when it hits in a few weeks' time.". Everywhere else i read, seems to think its due in Q3 or slightly before tho. So 4 months rather than a few weeks.

Thanks in advance.

[JumpingJack]

Anyone able to answer my questions please? They seem to have got lost a few pages back now. Here they are again:

1) Is the Conroe chip 64bit? --- If not, maybe AMD chips will have an advantage once Vista becomes widespread?

2) When is conroe due? Your article said, "It's abundantly clear that Conroe will be the fastest CPU in town when it hits in a few weeks' time.". Everywhere else i read, seems to think its due in Q3 or slightly before tho. So 4 months rather than a few weeks.

Thanks in advance.

1) Yes Core 2 (conroe) is a 64 bit CPU, all Core 2 and future Intel CPUs will be 64 bit. Core 2 for the laptop (called merom) will also be 64 bit enabled.

2) Conroe will arrive sometime in the first week or two of July, Q3 starts at the end of June. Do not be surprised if the announced date is pulled in. Not 4 months, more like 6 or 7 weeks out.

[keta]

its funny how people think q3 is like 4 months away this year has gone insanely fast , but q4 in just over a week is 4 months away , soon before u no it , it will b jan 2007 vista will b out and we will all b talkin about whos got the best quad core kentsfields or k8l , personaly i dont care which one is betta cuse im buying a conroe soon and the next processor i get will b a 8 core , soo ill skip the quad core cuse 8 cores aint far afta it.

and BladeRnR , there is a gigabyte 965P mobo coming out soon , its a top of hte line sli mobo that supports conroe , forget 975x , also 965G is probly the best conroe mobo , and ill b going for a ati xpress 3200 asus 965G , all those mobos will be out mid june to mid early july

[BladeRnR]

its funny how people think q3 is like 4 months away this year has gone insanely fast , but q4 in just over a week is 4 months away , soon before u no it , it will b jan 2007 vista will b out and we will all b talkin about whos got the best quad core kentsfields or k8l , personaly i dont care which one is betta cuse im buying a conroe soon and the next processor i get will b a 8 core , soo ill skip the quad core cuse 8 cores aint far afta it.

and BladeRnR , there is a gigabyte 965P mobo coming out soon , its a top of hte line sli mobo that supports conroe , forget 975x , also 965G is probly the best conroe mobo , and ill b going for a ati xpress 3200 asus 965G , all those mobos will be out mid june to mid early july

Thanks for the info Keta. Personally I don't like Gigabyte - ASUS or Abit for me

If Intel intends using Conroe in the new iteration of the Apple PowerPC then it's only logical the 965G will be in there to. I only hope it's an SLi board or there is an upgrade option for one!

Cheers

[acrobat]

1) Yes Core 2 (conroe) is a 64 bit CPU, all Core 2 and future Intel CPUs will be 64 bit. Core 2 for the laptop (called merom) will also be 64 bit enabled.

2) Conroe will arrive sometime in the first week or two of July, Q3 starts at the end of June. Do not be surprised if the announced date is pulled in. Not 4 months, more like 6 or 7 weeks out.

Wow cool thanks. I dont see AMD standing a chance against it then, until their new chip.

As for Q3, i read it was the end of Q3 which is why i thought 4 months. And also saw sites saying september which is about 4 months. But more recently ive seen end of june/start of july. If its around then, i will be really happy because i want to upgrade as soon as possible, and not waiting for the Conroe would be mad it seems.