I am a little skeptical about Conroe

[Duvie]

Originally posted by: Viditor

Originally posted by: ZOXXO

Originally posted by: Duvie
Dont forget business reports a few weeks back as Intel was announcing some less then rosy news...It seems they are having an issue with a large supply of chips in the warehouse...not being sold...They stated this amount was more then waht was needed for general warranty replacement...

...snip...

Not smart business sense...

Link?

Here's one, though you could do a Google search for "Intel warns" and find many more...

Thanks...I knew you would have been following this as well....

 

[imported_Questar]

1. who or what you work for?

I'm not comfortable giving out the details, but think healthcare. Really big heathcare.

2. why don't you use SIPP or CSIP programs like most businesses?
3. What exactly do you "qualify" on a desktop business platform that takes over a year??

We currently have just under 600 applications that have to be tested when a new hardware platform comes in. With three enginners, and eight airlock technicians it can take a while. Then you have all the infrastructure apps (identity management, integration apps, networks, web, systems mgmt., AD, etc).

You'd be amazed at some of the imcompatabilites that can crop up. I've seen a network driver that was .01 revision newer solve an application crash issue.

Then of course you need to build your base OS image and make sure it is very solid. When you're deploying 250 pc's a day - every day - it has to be rock solid. The costs of fixing something after it's out in the field is just huge.

 

[Viditor]

Originally posted by: Questar

1. who or what you work for?

I'm not comfortable giving out the details, but think healthcare. Really big heathcare.

Fair enough...

2. why don't you use SIPP or CSIP programs like most businesses?
3. What exactly do you "qualify" on a desktop business platform that takes over a year??

We currently have just under 600 applications that have to be tested when a new hardware platform comes in. With three enginners, and eight airlock technicians it can take a while. Then you have all the infrastructure apps (identity management, integration apps, networks, web, systems mgmt., AD, etc).

You'd be amazed at some of the imcompatabilites that can crop up. I've seen a network driver that was .01 revision newer solve an application crash issue.

Then of course you need to build your base OS image and make sure it is very solid. When you're deploying 250 pc's a day - every day - it has to be rock solid. The costs of fixing something after it's out in the field is just huge.

I do consulting for 12 Fortune 100 companies (GE is one of them), and none of them take more than 4 months to qualify a desktop business system (most are usually a month).
However, almost all of them are using SIPP or CSIP qualified platforms. This eliminates problems like early drivers and incompatabilities in all major software packages. I guess my questions are:
1. are all of your desktop systems required to be confirmed compatable with all 600 apps?
2. is there a reason that you aren't using guaranteed platforms from scratch?
3. are you talking about servers or desktops?

It seems to me that the IT design for your company could use some revamping if you're spending that many man-hours on qualifying mere desktops...

 

[Markfw]

Well, if he works for the same BIG healthcare company I do, their desktop support group are idiots. They want to push all 10k tnsnames entries to every desktop in the company. ! Half our software won't even work with that many entries. And my group only needs 15 entries. And you can;t even install software that is required to do your job, even if your department purchases it if it's not on the "approved" list. And that takes forever to get something on. Thats why I have my own personal PC to work on at home, its the only way I can get anything done.

 

[Childs]

Originally posted by: Viditor

Originally posted by: Childs
Intel Core Duo's at 2Ghz are faster than an 3800+ X2, and sit in between a X2 4200 and X2 4400. Hopefully this afternoon I can see how it does against an Opteron this afternoon. If the Core Duo is faster clock for clock than the X2, then I have no doubt that Conroe will be even better. Core Duo's arent necessarily cheap tho, so its hard to imagine the $316 price tag for something that bests anything AMD has by 20%. Doesnt seem to make much business sense, unless its old school gangsta vendetta sh1t.

Huh? Core Duo performs the same in most, worse in some, and better in very few benchmarks vs the X2 3800.
AT article

Based on my own testing. In Maya 7 the iMac 2Ghz is almost as fast as my Opteron 165 oc'd to 2.4Ghz. Less than 10% win for the opty average, typically 8%. Its reasonable to assume that my opty is 25% faster than a 3800 X2 in the same system. And on Synthetic benchmarks, the Core Duo is between the 4200 and 4400 X2. Its possible that the iMac is faster than a standard Core Duo, but to be honest I'm not even sure how valid AT's reviews are these days. Plus, that review you linked was a performance preview.

 

[kknd1967]

AT's review is using a pretty `lame' DDR2-533
Apple Power seriers and new Napa platforms should be usin DDR2-667 now

That should make some difference leap, 10% say, for Core duo

Originally posted by: Childs

Originally posted by: Viditor

Originally posted by: Childs
Intel Core Duo's at 2Ghz are faster than an 3800+ X2, and sit in between a X2 4200 and X2 4400. Hopefully this afternoon I can see how it does against an Opteron this afternoon. If the Core Duo is faster clock for clock than the X2, then I have no doubt that Conroe will be even better. Core Duo's arent necessarily cheap tho, so its hard to imagine the $316 price tag for something that bests anything AMD has by 20%. Doesnt seem to make much business sense, unless its old school gangsta vendetta sh1t.

Huh? Core Duo performs the same in most, worse in some, and better in very few benchmarks vs the X2 3800.
AT article

Based on my own testing. In Maya 7 the iMac 2Ghz is almost as fast as my Opteron 165 oc'd to 2.4Ghz. Less than 10% win for the opty average, typically 8%. Its reasonable to assume that my opty is 25% faster than a 3800 X2 in the same system. And on Synthetic benchmarks, the Core Duo is between the 4200 and 4400 X2. Its possible that the iMac is faster than a standard Core Duo, but to be honest I'm not even sure how valid AT's reviews are these days. Plus, that review you linked was a performance preview.

 

[Viditor]

Originally posted by: Childs

Originally posted by: Viditor

Originally posted by: Childs
Intel Core Duo's at 2Ghz are faster than an 3800+ X2, and sit in between a X2 4200 and X2 4400. Hopefully this afternoon I can see how it does against an Opteron this afternoon. If the Core Duo is faster clock for clock than the X2, then I have no doubt that Conroe will be even better. Core Duo's arent necessarily cheap tho, so its hard to imagine the $316 price tag for something that bests anything AMD has by 20%. Doesnt seem to make much business sense, unless its old school gangsta vendetta sh1t.

Huh? Core Duo performs the same in most, worse in some, and better in very few benchmarks vs the X2 3800.
AT article

Based on my own testing. In Maya 7 the iMac 2Ghz is almost as fast as my Opteron 165 oc'd to 2.4Ghz. Less than 10% win for the opty average, typically 8%. Its reasonable to assume that my opty is 25% faster than a 3800 X2 in the same system. And on Synthetic benchmarks, the Core Duo is between the 4200 and 4400 X2. Its possible that the iMac is faster than a standard Core Duo, but to be honest I'm not even sure how valid AT's reviews are these days. Plus, that review you linked was a performance preview.

A LOT of problems with this comparison...it's about as roundabout as possible.

1. Which version Maya on the Mac, and which one on the PC
2. Have you checked for processor usage on the PC? (is the patch and bios installed properly)
3. What Ram and how much are you running on either system?
4. Opty OC to 2.4 shouldn't be 25% faster than the 3800, more like 10%...but even then it will depend on the chip and ram (no 2 overclocks are identical).
5. What dividers did you use for the OC?

The preview was the review just before release and used production chips. But if you can find some review that shows otherwise (except THG of course), then I'd be very happy to see it! I have more respect for Anand's reviews than any other reviewer...that said, even Anand makes mistakes and if you've found one please share!

Edit: Another thing I just thought of...how many threads are running in the background on the PC? Are you running both systems at the same resolution?

 

[Absolute0]

"4. Opty OC to 2.4 shouldn't be 25% faster than the 3800, more like 10%..."

That's the most retarded thing i've heard all day. 3800+ is 2 Ghz, 10% added to that is 2.2 Ghz. 25% added to it is 2.5 Ghz, and since he's running at 2.4 Ghz with double the cache, a 2.5 Ghz X2 is spot on, and 25% is the correct number, 10% is frigging bogus.

And the Core-Duo benches, i've seen the results, undoubtedly faster than AMD64 X2 on a clock per clock basis, but not by a large margin. Merom will replace it as a mobile processor, adding Conroe-like performance to the mobile arena.

 

[Apocalypse X]

Originally posted by: Quinton McLeod

Originally posted by: Accord99

Originally posted by: Quinton McLeod

Is it me or... Does this look suspecious?
http://www.xtremesystems.org/forums/showthread.php?t=95021&page=1

Take a good look at those pics. Why is it that the Conroe has a voltage over 3.0 volts? If the Conroe's TDP is suppose to be lower than AMD processors, then why is the voltage so high? I mean, for AMD processors, anything over 1.6 volts would be considered meltdown.

It's because, if you read the thread, that cpu-z does not properly detect the voltage. Everest does detect the proper voltage of 1.2v. It was also noted that the Conroe chip is passively cooled.

Link?

Earth to Quinton...its time to go beddybyeeeee

 

[Viditor]

Originally posted by: Absolute0
"4. Opty OC to 2.4 shouldn't be 25% faster than the 3800, more like 10%..."

That's the most retarded thing i've heard all day. 3800+ is 2 Ghz, 10% added to that is 2.2 Ghz. 25% added to it is 2.5 Ghz, and since he's running at 2.4 Ghz with double the cache, a 2.5 Ghz X2 is spot on, and 25% is the correct number, 10% is frigging bogus.

And the Core-Duo benches, i've seen the results, undoubtedly faster than AMD64 X2 on a clock per clock basis, but not by a large margin. Merom will replace it as a mobile processor, adding Conroe-like performance to the mobile arena.

You need to get out more...
Overclocks don't represent a linear performance improvement (far from it).
Take a look at the following link.

Note that the X2 4200 OC from 2.2 up to 2.7 (a 22% OC) only gains it a 5-20% advantage...however the 20% advantage is very rare in the tests (most all are in the 5-10% range). While many of the inexperienced believe that a 25% overclock gets them a 25% system speed increase, that never is the real case...

As to your testimony on Core-Duo, if you would please post some actual benches (i.e. facts) it would be greatly appreciated!

 

[designit]

Originally posted by: Absolute0
"4. Opty OC to 2.4 shouldn't be 25% faster than the 3800, more like 10%..."

That's the most retarded thing i've heard all day. 3800+ is 2 Ghz, 10% added to that is 2.2 Ghz. 25% added to it is 2.5 Ghz, and since he's running at 2.4 Ghz with double the cache, a 2.5 Ghz X2 is spot on, and 25% is the correct number, 10% is frigging bogus.

And the Core-Duo benches, i've seen the results, undoubtedly faster than AMD64 X2 on a clock per clock basis, but not by a large margin. Merom will replace it as a mobile processor, adding Conroe-like performance to the mobile arena.

not so fast.
XS VW who tested conroe(BTW he is an Intel's enthus) shows that conroe does less than 700mb/s winrar. My A64 Venice @ 2.7ghz does 749mb/s. I think conroe is fast in integer calculations as SPi indicates, but not so fast in other areas. And we wouldn?t know if this aggressive disambiguation of memory that has 3% error probability is
going to cause any alarm or not. SPi is very simple arithmetic calculation and easily predictable by core's branch prediction. When retail version of both Conroe and Socket M2 hit the market and un-bias comparison/reviews such as one by AT is conducted, only then, I believe it. I Do not trust any test coming out of XS Intel's forum.

 

[Absolute0]

you don't trust Victor Wang? Notice his name is in blue, he's XIP, basically if you don't trust him, it is because of misguided reasons...

so... you're RimRam in that thread.

 

[Childs]

A LOT of problems with this comparison...it's about as roundabout as possible.

1. Which version Maya on the Mac, and which one on the PC
2. Have you checked for processor usage on the PC? (is the patch and bios installed properly)
3. What Ram and how much are you running on either system?
4. Opty OC to 2.4 shouldn't be 25% faster than the 3800, more like 10%...but even then it will depend on the chip and ram (no 2 overclocks are identical).
5. What dividers did you use for the OC?

The preview was the review just before release and used production chips. But if you can find some review that shows otherwise (except THG of course), then I'd be very happy to see it! I have more respect for Anand's reviews than any other reviewer...that said, even Anand makes mistakes and if you've found one please share!

Edit: Another thing I just thought of...how many threads are running in the background on the PC? Are you running both systems at the same resolution?

1. The iMac is running XP w/SP2, same as the PC. Both ran the exact same version of Maya, which was 7 Complete.
2. CPU usage is 100% on both. If you dont render Maya scenes you may not understand this, but Maya can use up to 4 threads. If I had dual xeons, I would render with 4 threads, and dual core systems with 2. The goal is to see what the best performance is, so it would make no sense to run at anything less than 100% CPU usage. Patch installed, latest bios on KN8 (v15).
3. 1GB 400Mhz DDR in Opteron, 1GB DDR in iMac. Exact speed of iMac memory escapes me at the moment, but I can find out on Monday.
4. I'll stick by the 25% until I see different. Maybe after breakfast I'll downclock the Opteron to 1.8Ghz and bench again, as that would be inline with AMD's rating scheme of 200Mhz vs 1MB L2 cache. I'm fairly certain its a fair approximation, +-5% at most.
5. Dividers 4X HTT, 333Mhz.
6. Both times Maya was the only thing running after a reboot. I was evaluating the Maya performance on the PC versus Mac, so I wanted the best performance.

 

[Duvie]

Childs,

I have a copy of Maya (version 6 I think)...I would like to test run it on my dual 270's...Do you have a project you can send me to render??? For comparison...

 

[designit]

Originally posted by: Absolute0
you don't trust Victor Wang? Notice his name is in blue, he's XIP, basically if you don't trust him, it is because of misguided reasons...

so... you're RimRam in that thread.

What kind of asenine asumption is that?
I said he is Intel's enthusiast. how do you read "no trust" in that sentence?
If you didn?t understand why I eluded to that Let me clarify it. He would have done everything possible to correct any error concerning conroe's winrar test. another word, conroe doing <700 winrar is not a mistake and VW is not cooking it (because he might be AMD biased). Is this clear or do I need to break it down more for you?
Gee, it is hard to believe how people twist the meaning of a sentence.

 

[Childs]

OK, here are my results:

1.8Ghz

scene1 1:34
scene2 1:05
scene3 1:10
scene4 9:10

2Ghz

scene1 1:24
scene2 0:58
scene3 1:02
scene4 8:17

2.4Ghz

scene1 1:11
scene2 0:49
scene3 0:52
scene4 6:54

So 2.4Ghz is about 25% faster than the 1.8Ghz. In 3D rendering apps, a dual core 1.8 with 1MB is a tad faster than a dual core 2Ghz with 512K. All components in the test were the same, simply changed fsb. iMac Core Duo at 2Ghz rendered scene1 in 1:21. I didn't do the others as that machine is at work, but if anyone really wants to know I can do it on Monday.

Duvie, YGPM.

 

[Tangerines]

Originally posted by: designit

Originally posted by: Absolute0
"4. Opty OC to 2.4 shouldn't be 25% faster than the 3800, more like 10%..."

That's the most retarded thing i've heard all day. 3800+ is 2 Ghz, 10% added to that is 2.2 Ghz. 25% added to it is 2.5 Ghz, and since he's running at 2.4 Ghz with double the cache, a 2.5 Ghz X2 is spot on, and 25% is the correct number, 10% is frigging bogus.

And the Core-Duo benches, i've seen the results, undoubtedly faster than AMD64 X2 on a clock per clock basis, but not by a large margin. Merom will replace it as a mobile processor, adding Conroe-like performance to the mobile arena.

not so fast.
XS VW who tested conroe(BTW he is an Intel's enthus) shows that conroe does less than 700mb/s winrar. My A64 Venice @ 2.7ghz does 749mb/s. I think conroe is fast in integer calculations as SPi indicates, but not so fast in other areas. And we wouldn?t know if this aggressive disambiguation of memory that has 3% error probability is
going to cause any alarm or not. SPi is very simple arithmetic calculation and easily predictable by core's branch prediction. When retail version of both Conroe and Socket M2 hit the market and un-bias comparison/reviews such as one by AT is conducted, only then, I believe it. I Do not trust any test coming out of XS Intel's forum.

I'd assume that the lower score had to do with the system's RAM deficiency more than anything.

 

[Duvie]

Originally posted by: Childs
OK, here are my results:

1.8Ghz

scene1 1:34
scene2 1:05
scene3 1:10
scene4 9:10

2Ghz

scene1 1:24
scene2 0:58
scene3 1:02
scene4 8:17

2.4Ghz

scene1 1:11
scene2 0:49
scene3 0:52
scene4 6:54

So 2.4Ghz is about 25% faster than the 1.8Ghz. In 3D rendering apps, a dual core 1.8 with 1MB is a tad faster than a dual core 2Ghz with 512K. All components in the test were the same, simply changed fsb. iMac Core Duo at 2Ghz rendered scene1 in 1:21. I didn't do the others as that machine is at work, but if anyone really wants to know I can do it on Monday.

Duvie, YGPM.

Thanks Childs...

I will be out of town for 3 days but when I get back I will run it....

I can tell you that in CAD apps it does scale quite well in terms of mhz...I did not get 100% gain going to dual core but when I change mhz thru ocing the scaling was pretty close to linear in these types of apps....

cache has meant very little though in my testing....

http://forums.anandtech.com/messageview...hreadid=1531414&enterthread=y&arctab=y
http://forums.anandtech.com/messageview...hreadid=1637764&enterthread=y&arctab=y

 

[Viditor]

Originally posted by: Childs
OK, here are my results:

1.8Ghz

scene1 1:34
scene2 1:05
scene3 1:10
scene4 9:10

2Ghz

scene1 1:24
scene2 0:58
scene3 1:02
scene4 8:17

2.4Ghz

scene1 1:11
scene2 0:49
scene3 0:52
scene4 6:54

So 2.4Ghz is about 25% faster than the 1.8Ghz. In 3D rendering apps, a dual core 1.8 with 1MB is a tad faster than a dual core 2Ghz with 512K. All components in the test were the same, simply changed fsb. iMac Core Duo at 2Ghz rendered scene1 in 1:21. I didn't do the others as that machine is at work, but if anyone really wants to know I can do it on Monday.

Duvie, YGPM.

Good info! It surprises me mainly because in the link I posted, they were only getting a 16% increase on a 22% overclock for Maya in Specview Perf...
Anyone have a speculation on what the difference might be?
Childs...how many times did you run the bench?

I did note that of all the tests, Maya allows for one of the largest increases in performance from overclocking.

Edit...posted while Duvie posted. That must explain it then, rendering seems to make very efficient use of the extra MHz. Thank you very much for the info and the effort gentlemen!

Edit 2: Duvie, I went through your excellent post on bench results, and I come up with an average increase of 37.88% for an overclock of 48.1%
What strikes me as odd is that many of your results show a higher than linear increase from the OC...for instance, ScienceMark showed a 55.6% improvement on a 48.1% overclock. Any speculation as to how that happens (from anyone) is appreciated!

 

[Childs]

I ran each test 2 or 3 times, depending on whether or not the second run varied from the first. When it did vary, it was by 1 second, and the same on the 3rd. Remeber, I stated an opteron oc'd to 2.4Ghz was 25% faster than a 3800+ X2 in Maya. I dont have my X2 anymore, but that was what I remembered when I benched it.

Anyways, Conroe should do what everyone thinks so far, since the Core Duo is the same or slightly better than X2's at the same clock speed. Should be great value.

 

[Corporate Thug]

guys, i dont want to start a new thread to ask this so i will ask it here: when is conroe supposed to be released? i hear some people saying september and some people saying july

 

[Viditor]

Originally posted by: Childs


Anyways, Conroe should do what everyone thinks so far, since the Core Duo is the same or slightly better than X2's at the same clock speed. Should be great value.

You really can't say that the Core Duo is better than X2 from what you've posted, only that it's better in Maya...big difference.
A caveat here...The Conroe and Core Duo are 2 different architectures from the same basic principals. That said, I agree that Conroe will be better at Maya...
Actually, it's looking like Intel is coming out with another "Northwood win" at last. The only questions left now are:
1. What/when will availability be like on Conroe
2. When will K8L be out to strike back
3. When will AM2 hit 65nm for the desktop (servers appear to be headed for a first release in Q4 of this year)

 

[imported_Questar]

1. are all of your desktop systems required to be confirmed compatable with all 600 apps?

Yes. Those are apps that run the business. SIPP or CSIP programs don't help here as a large portion of those apps are written in house (we have 880 programmers).

2. is there a reason that you aren't using guaranteed platforms from scratch?

We do to some extent, but remember a guarantee is not a warranty.

3. are you talking about servers or desktops?

Both

It seems to me that the IT design for your company could use some revamping if you're spending that many man-hours on qualifying mere desktops...

I agree. But you know how it goes in business, you pick your battles.

 

[DrMrLordX]

Originally posted by: Corporate Thug
guys, i dont want to start a new thread to ask this so i will ask it here: when is conroe supposed to be released? i hear some people saying september and some people saying july

Last I heard, it was somewhere between June/July. Conroe should see the light of day at about the same time AM2 hits the market.

 

[Childs]

Originally posted by: Viditor

Originally posted by: Childs


Anyways, Conroe should do what everyone thinks so far, since the Core Duo is the same or slightly better than X2's at the same clock speed. Should be great value.

You really can't say that the Core Duo is better than X2 from what you've posted, only that it's better in Maya...big difference.
A caveat here...The Conroe and Core Duo are 2 different architectures from the same basic principals. That said, I agree that Conroe will be better at Maya...
Actually, it's looking like Intel is coming out with another "Northwood win" at last. The only questions left now are:
1. What/when will availability be like on Conroe
2. When will K8L be out to strike back
3. When will AM2 hit 65nm for the desktop (servers appear to be headed for a first release in Q4 of this year)

On synthetic benchmarks (Sandra) it beat the 3800 X2 outright, and was in between the 4200 and 4400. If I still had the X2 I would do more benches to state it with certainty, but I have no reason to think otherwise ATM. Maybe I'll run some spec benches later today. The video card in the iMac is mediocre, so I'll run any test that tests CPU only.

EDIT: DOH! I thought it was free! Well scratch the spec benches.