Collection of Conroe Data. (Core 2 Duo and Core 2 Extreme!)

[RichPLS]

Christmas is looking brighter and better...

 

[RichPLS]

Christmas is looking brighter and better...

 

[RichPLS]

Christmas is looking brighter and better...

 

[ltcommander_data]

For interest, Rahul Sood who hasn't always been optimistic about Conroe's performance, in fact he was one who exaggerated the IDF BIOS issue, is now on side.

http://voodoopc.blogspot.com/2006/06/rumor-mill-ati-amd-does-it-make-sense.html

In the meantime people have been questioning whether or not I've "switched camps" - and I think you need to realize that I was never in one particular camp. Voodoo has always supported the best technology, that is why up until now we sell 100% AMD on the desktop and workstation, and about 50/50 AMD to Intel on notebooks. People can't expect Intel to remain in this position forever and as such we have been eagerly awaiting the launch of Conroe. I have been testing it, and as I said before it's *!%#ing fast. I honestly believe Intel Conroe is going to revive our industry and kick things into high gear.

He even goes so far as to say that now may be a good time to buy Intel stock. (He also saids that AMD stock is a good option too). Sharikou then jumps at the suggestion that Sood would suggest buying Intel stock. You notice that Sharikou no longer downplays Conroe's performance as strongly, but instead argues based on Netburst inventory and AMD's planned cores. Even Sharikou has admitted that Conroe's better performance is not fake.

http://sharikou.blogspot.com/2006/06/conroe-getting-really-close-to-be.html

Except a few simplistic tests, the E6800's lead over FX62 is around 10% in majority of the tests.

This stepping 5 will probably be the shipping version, since Intel really aught to get stock piling now to alleviate supply concerns. It'd be interesting to see whether this lower voltage will allow the 2.93GHz model to fall into the 65W TDP. They probably wouldn't label it as such though since they need to X to indicate its an Extreme Edition.

 

[iterations]

June 16 2006: Well they said it could be done and now we have proof. Conroe can hit 4GHz on Air cooling alone! Denny has gotten his hands on a new Stepping-5 (likely retail sample) of an E6700 Core 2 Duo and paired it with his Tower 120 Air Cooler.

Link: http://www.xtremesystems.org/forums/showthread.php?t=103381

Discussion thread: http://forumz.tomshardware.com/hardware/Retail-Stepping-Conroe-hits-4GHz-air-cooling-ftopict188251.html

13 Second SuperPi_1M run on air cooling...

 

[ltcommander_data]

I have most of this in your other Stepping 5 thread, but since there are some links I'd thought I'd add it to the official log.

Needless to say, that is an amazing result. I guess it just means that Intel has plenty of clocking room left, which was what I was worried about. This should mean that they still have some maneuvering room in case there are issues with 45nm. I still wish Intel wasn't so conservative with their Extreme Edition release at only 2.93GHz.

I'm sure you also noticed the other thread at Xtreme Systems.

http://www.xtremesystems.org/forums/showthread.php?t=103390

They've reached a 500MHz base system clock rate for a 2001MHz FSB. Nearly double what the shipping FSB will be. They're dumping in plenty of voltage and had to drop a multiplier on the processor but it's still impressive. After this you'd think a 1333MHz FSB would be a pittance.

VR-Zone also had an in-depth look at the Gigabyte GA-965P-DQ6 which claims to be quad core compatible.

http://www.vr-zone.com/?i=3735

With the 965 chipset and the 12 phase power (more of an advertising feature really since Conroe isn't that needy) it should be a great overclocking motherboard. Probably quite competitive with the RD600 in that respect. The only thing holding it back from being a full enthousiast motherboard is the lack of Crossfire support. It's not really that much of an issue now though with the 7950GX2.

 

[iterations]

And the feat has now been duplicated by Vapor, who also has a Stepping-5 Core 2 Duo E6700

This time it is in a case, and is using a different heatsink (Typoon).

Link: http://www.xtremesystems.org/forums/showpost.php?p=1518585&postcount=53

Cheers!

P.S. And before anyone comments "it is water cooling!!!11~", in the link above, READ THE DAMN THREAD. The water cooling is only on his graphics cards, not the CPU. The CPU is air cooled only.

 

[lbax]

It is my belief that in the context of recent dual core cpus, Intel has only offered Hyper Threading as a premium feature on it's top of the line EE products. Is HT being carried forward to the Conroe architecture, or is it a Netburst specific feature/function? Does anyone have any links to whitepaper discussion of this function as it relates to Conroe?

 

[mjp1618]

At least the initial Core products will not have Hyperthreading.

http://www.theregister.co.uk/2005/08/23/intel_next_gen_architecture/

 

[lbax]

Yes, I have seen several third party references that Initial Core products will not have Hyperthreading. They leave me asking the question "is Hyperthreading going to be implemented in the Core architecture"? I was hoping that some of our regular posting "insiders" could provide more current insight into the questions surrounding multi-task &/or multi-thread support within Core. Specifically, is HT part of the Core roadmap or is it germane only to Netburst?

 

[Kougar]

While I'm not exactly in the know, from everything I have read and seen to date says the Core uArch was never designed with Hyper-Threading in mind. One of the reasons given is that HT will increase the TDP/power draw of a processors that uses it, and I have noticed this significantly with my own Northwood P4 that has HT. Vcore draw readings drop by ~.08v and temps drop 5c or 6c at 3.42ghz.

However I have come across at least two sites that were either hinting or had written supposition to the effect that Intel might yet add HT support to the 3.2ghz EE Conroe. Considering that from what I've read Intel's results with "Core" have been better than they were expecting, they very well may have the TDP headroom to add it back in. Even Ars Technica's article stated the "Core" uArch was much better suited for HT use than even Netburst's own design, and could better use such a feature.

Of course one of the same sites that mentioned Conroe might have HT also stated Conroe would be delayed from it's July 23rd launch until September and no one else has mentioned this... so I can't vouch for the credibility. Quote is from the TweatTown article in question:

Hyper Threading technology stated to be disabled for all processors at the moment but we could see Intel releasing a new Core 2 Extreme processor with HT enabled at clock speeds over 3GHz.

 

[ltcommander_data]

Does anyone have any links to whitepaper discussion of this function as it relates to Conroe?

I don't know of any white paper discussions, but I can point you to my own feelings on HT and Conroe. I've actually brought this up a number of times.

http://forumz.tomshardware.com/hardware/modules.php?name=Forums&file=viewtopic&t=178335&highlight=

Basically, my feeling is that HT would work very well with the Core architecture because of it's wider 4-issue design and extra execution units. This should really help it sustain 2 threads. A major issue with Netburst and HT was that it divided up the system resources such as buffers and cache which had a impact on the performance of each thread. This especially became apparent while 1 thread stalled in Replay mode while still hogging resources. The Core architecture would avoid a lot of the resource issues because buffers and such are all already bigger as part of the 4 issue design. In fact the Reorder Buffer has actually more than doubled from Yonah even though the change was only from 3 issue to 4. Cache trashing between logical cores in Netburst would also not occur in Core because the shared L2 logic already exists and just needs to load balance for 4 logical cores instead of the 2.

My feelings is that if HT were to be introduced in Core the best time to do it would be when they transistion to 45nm. The increase in cache from 4MB to 6MB will ensure that each logical core has more resources while the shrink will provide room to alleviate the additional power and thermal loads. Even then, the concern that HT generates excessive heat is probably exaggerated. Heat in and of itself isn't bad, it's just when heat is generated without doing work which was the case when a thread was stuck in Replay in Netburst. Core doesn't have a Replay loop. Core's more power efficient architecture will also help.

Given the statements from Intel it's quite possible that there is some preliminary work with HT on Core. Initially they avoided saying whether HT was dead, and I believe they've always maintained that it was a great idea that has just been postponed until it's required. I would be disappointed if they didn't tinker around with different features that just haven't been implemented yet. HT was supposedly in Netburst since the first Pentium 4s even though it wasn't activated. The L1-L1 bridge in Core also appears to be a feature that was tried but hasn't been activated.

The question is whether it will be. Most likely not. Recently Intel has taken great lengths to differentiate Core 2 and Netburst so it wouldn't be wise to bring back Netburst until the memory has dulled a bit. Still it has shown benefits in server applications where multithreading is more common so if any product were to receive it it would be Xeons. Given that both Intel and AMD feel that the rush is to quad core, than a slower pace to 8 core and even slower beyond, HT would be a way for Intel to market higher threading without actually spending the extra transistors. That was the original plan for HT anyways.

 

[ltcommander_data]

I don't know where TweakTown gets its information, but as far as I know there are no official plans to have HT support in Conroe. Even if HT were to be added, it would not be to the 3.2GHz Extreme Edition since that'd be far too close to the quad core Kentsfield. As I said in my earlier post, the best time would be to a true (single die) quad core 45nm chip in order to get 8 cores.

 

[Kougar]

Nice posts ltcmder Data!

I don't think this thread was pointed out yet, but I didn't check to confirm... http://forums.guru3d.com/showthread.php?t=185555 An ES sample E6600 and an Opteron 165, both compared at 3ghz with pretty similar configs. Only difference is the RAM and motherboard.

 

[lbax]

THANK YOU! Great post.

 

[iterations]

HKEPC got the latest and greatest from both AMD and Intel and did a full range of tests, see below.

All tests performed with the same graphics cards, hard drives, memory, memory timings, and OS:

With Crossfire x1900s:

Cheers!

 

[iterations]

Nah, not really frustrated, but it is amusing how some fanboys of a particular flavor insist on the "head in sand" response to hard data.

There is one thing I'd really like to point out in the data I just posted. HKEPC did real temperature and energy consumption testing with external sensors. Obviously Core 2 Extreme kicked butt performance wise, but really check these out:

Core 2 Extreme uses FAR less energy at idle and at full load. And it runs are FAR lower temperatures, again both at idle and full load.

That is very impressive to me. Not surprising, as that is what Intel was aiming for, but impressive none the less.

 

[Action_Man]

Impressive stuff but what are the specs of the rigs?

 

[Artmic]

Can anyone say for certain that a Woodcrest 3Ghz CPU is to cost $851 ? I've seen the price online on some sites, and it dont' make sense to me, why is it cheaper than the 2.93Ghz Conre ? with the crappier Bus speed?

 

[iterations]

To Action_Man: (sorry I forgot to post the specs)

Intel Core 2 Extreme X6800 2.93GHz/4MB
MSI 975X-Platinum (i975X + ICH7R + VRM11)

AMD Athlon 64 FX 62 (AM2) 2.8GHz/1MB x2
Foxconn C51XEM2AA (nForce 590SLi)

Memory for both: Corsair DDR2-800 1GB x 2 (CL5-5-5-15)
Graphics for both: Gigabye Geforce 7800GTX 256MB (430MHz/1.2Ghz
OS For both: Windows XP Professional w/ SP2
Ambient lab temprerature for both: 26C
Cooler for both: Gigabyte G-Power lite @ 2000RPM
Temperature Sensor for both: Type K Thermo


To Artmic:

Woodcrest needs a server class motherboard which adds alot of cost to the system overall system, and it is a different socket type (not LGA775 like Conroe). Core 2 Extreme has an unlocked multiplier so you can overclock it very aggressively. It is an enthusiast part, where the 3.0GHz Woodcrest part is a high end server part. Intel is just providing very good pricing on these parts since they can afford to with their 65nm fab process.

Cheers!

 

[Artmic]

To Action_Man: (sorry I forgot to post the specs)

To Artmic:

Woodcrest needs a server class motherboard which adds alot of cost to the system overall system, and it is a different socket type (not LGA775 like Conroe). Core 2 Extreme has an unlocked multiplier so you can overclock it very aggressively. It is an enthusiast part, where the 3.0GHz Woodcrest part is a high end server part. Intel is just providing very good pricing on these parts since they can afford to with their 65nm fab process.

Cheers!

Thanks for the info, it makes sense, i forgot about the other parts of the equation, the motherboard will probably cost $500+ and the memory will also be more expensive probably, add to that the Extreme edition you can OC like crazy i guess and you got the right price lol

I am just curious how much faster a comparably clocked woodcrest is against a Conroe with its memory/bus speed.

 

[iterations]

I am just curious how much faster a comparably clocked woodcrest is against a Conroe with its memory/bus speed.

Now that is a very interesting question. I'm curious to see this as well, even if they are meant for different market segments. I'm sure we'll see this in July sometime.

 

[Artmic]

I am just curious how much faster a comparably clocked woodcrest is against a Conroe with its memory/bus speed.

Now that is a very interesting question. I'm curious to see this as well, even if they are meant for different market segments. I'm sure we'll see this in July sometime.

Very True, i can't wait, i haven't been this excited about Intel since the P3 days........

 

[1Tanker]

HKEPC got the latest and greatest from both AMD and Intel and did a full range of tests, see below.

All tests performed with the same graphics cards, hard drives, memory, memory timings, and OS:

I like this review.

http://www.tbreak.com/reviews/article.php?cat=cpu&id=456&pagenumber=2

Tighter memory timings help AMD Fanboys hold out hope for AM2 BUT, these numbers bode VERY well for Conroe, as it shows that you could get a cheapo DELL, for instance with higher latency(cheap) RAM...and still get world-class performance....Not the case with AM2. 😀

 

[Kougar]

Madshrimps posted a max overclocking on air article for the X6800 here.

It was only a revision 4, but they still reached 3.3ghz before any change to the Vcore was needed, and then 3.75ghz on air. Article includes some light benches.