To C2Q or not to C2Q, that is the question. (IMHO)

I have received quite a few questions related to the Quad CPU's with our systems. This thread is to answer those questions and publicly post my advice on the subject. While the Quad cores are certainly powerful there are some inherent limitations.

When upgrading users are trying to decide between the extreme versions of CPU's of either the x9100 C2D or quad variants of the q9200 or qx9300. For the purpose of this post I have to go by my respective upgraded P7805u and cooling system in relation to the various attainable parameters.

The first consideration is what clock speeds can be reached with the respective CPU's. With a x9100 a clock of 3.7 GHz is usable for about everything other than Prime95 for extended runs without overheating. The quad cores are able to run at 2.93 with the same results. Under low to normal usage the x9100 could obtain 4.0 GHz and the quad core 3.2 GHz. For sustained walk away 24 hour runs of high load such as prime95 and no heat issues you can run 3.45 GHz for the x9100 and 2.66 GHz for the quad core. The x9100 can obtain much higher overall core clocks for the related stability and cooling solution.

Now with single or dual thread applications, the majority of what is out there, the OS pools the thread(s) and off loads it to all respective available cores by interlacing the load. So on a C2D 1 thread could load each core to 50% where on a quad core the same thread could load each core to 25%. No one thread though will execute faster than the overall clock speed. Here a C2D shows a distinct advantage from the higher attainable clock speeds. That however is not the end of the story. In order to load the two physical silicon cores to interlace the single thread the individual cores caches must load program and data etc. The same for outputting data to ram. This duplication of data etc will give a performance hit to the thread of up to 7% at the same clock speed to the single silicon cored C2D.

This can be shown on a C2Q by taking SuperPi single thread and setting affinity to core 0 & 1. Setting the affinity will give you up to a 7% increase in performance at the same CPU clock by using just the one physical core and cache structure. As a side note it should be realized with each CPU duplicating the data in cache on pooled threads that the 6MB L2 of each core is not acting then as a true total 12MB cache.

Now on a quad optimized application without thread pooling the quad will really shine. These applications are few and far between and also are usually specialized. They do exist though and in the future more may come to light.

Those with the new P79 main boards, and possibly present P79xx owners, can upgrade to x9100 or quad variants. Most P78xx users, other than systems originally quad core equipped, are limited to a C2D only. Present P78xx with a q9000 CPU though should be able to upgrade with a x9100 or C2Q extreme CPU's.

Recommendations;

1.) x9100
Unless you are seeing the CPU flooded often at 100% usage the C2D is the better choice. Most gaming applications, where CPU usage is hidden from view, are better optimized for the C2D. Without the thread pooling issue of the quad core the C2D will generally feel much snappier and more responsive. The C2D also tends to have a better heat profile for the P78 and P79 platforms cooling solutions.

2.) q9200 & qx9300
If you tend to run multithread optimized programs or extreme multitasking where lots of programs can utilize the thread pooling the quad variants are for you. Clock for clock you may scratch you head as to why it stumbles ever so slightly. Over all though the Quad CPU's are proverbial tanks. Nothing, at least that I've found, stops them. If you are coming though from a x9100 over 50% of the time under general usage you actually may be somewhat disappointed. You will lose allot of the x9100 snap, clock for clock the quad just does not match the C2D and of course you cannot even dream of reaching the total x9100 cores clock speeds.

Notes;

Now for synthetics 95% of the time or better the C2Qs will win hands down. For real world usage 95% of the time or better the C2D is the better choice. You then have to decide if you spend more time using the system than just bragging about it.

Now this thread in no way encroaches on the iCore. The newer CPUs alleviate the thread pooling issue by having a shared L3 with smart cache. Intel apparently learned from the C2Q's failings and without uncovering the C2Q's thread pooling issue and just fixed their original mistake. Just remember where C2Q's are concerned 12 MB cache is really 6MB duplicated in thread pooling unless the program is quad core thread optimized.

Arguably better thread pooling optimized specifically for C2Qs would place overall performance closer to the new quad iCores. Since these are at, or near, EOL do not expect this anytime soon. This is however a debate for another thread in another forum.

Conclusion;

Personally I plan to soon highly use the C2Qs multithread capability. Rather than build a desktop to proccess that function I am staying with the C2Q. I could care less about bragging rights or synthetics myself..............

 

Sounds like you need a qx9300 to test next.

 

LoL, I doubt the qx9300 will yield more than where I am at with the q9200. The quad core has been quite an education. I did research ways to enhance the performance knowing the structural limitations of the quad core and the dual silicon structure.

If M$ treated the CPU as two physical C2D CPU's in pooling the threads we would have better single thread performance but there is a trade off in pooling efficiency under multithreading and high CPU loads. There is always something to deal with.

What I need now is a i7-2920xm with a LED LCD 1920x1200 to play with……………..

 

Good job on this!

 

Got it, thanks, actually had to fix a few issues, sure there is more too.............

 

wow bro all I could say is amen...you really preached it.

I was about to go Quad with this laptop. But since you've made it plain as day. The only programs I use that will benefit are my video converting & my Cinema 4D.

 

Thanks.................

 

I'd take a 3ghz dual over a 2Ghz quad, but i'd take a 3ghz quad over a 4ghz dual.

I can run my quad at its floor voltage (1.05V) at 2.66ghz at which point its running cooler than a stock X9100 so I am surprised thats the most you could get out of it for heavy loading, I barely break 60C when gaming (shared heatsink) at that point with a massive 875/1100 overclock on my HD5730.

Also as we go forward and quads are used more we are seeing games like farcry 2 showing 20% FPS differences with a quad over dual.

 

I can normal run use 3.2 GHz with the Q9200. for most tasks out there today the x9100 kicks it to the curb. 3.2 GHz, unless the threads are programed for multiple core with affinity, has to pool the threads to each core. This causes a quad core to loose 7% of the speed from efficiency. So a single non optimized thread only runs about as fast as it would on a 2.933 GHz. CPU.

My q9200 runns 2.66 at 1.05 floor as well and under normal usage barely hits 60c. Run Prim95 4 core on it for an extended run, it will get hotter though.

Now agreed if the program is multi thread optimized you will get a boost, even more so on the icore structure as you don't have that 7% loss going between physical silicon cores. As games become multi thread optimized multi cores will show their brute force, but for now most are not and it will be a while before they are.

 

There quite a few multi-core friendly games out already and their number are only increasing, at least up to 4 threads.

 

but not yet that fully advantage a C2Q over a C2D as in both being extreme versions and overclocked.................

 

If a program is properly threaded for 3-4 cores then the quad will win when both are overclocked.

I was looking at a comparison between the E8600 (3.33ghz) and Q9650 (3ghz).

The only game to show a hit at the same frequency (E8400 vs Q9650) was left 4 dead.

 

Right, at these speeds the GPU's are usually the bottleneck. two thngs though.

1.) for the same clock the C2D is way cooler than the C2Q
2.) the C2D can be clock way higher than the C2Q

For this reason, even without efficiency loss of the C2Q in non optimized threads, the C2Q is a much snappier core. Eith will game comparibly at 3.0 GHz and above.

The idea here is why game the same, have a less snappy system that runs hotter, possibly spend more or even to do an upgrade that does nothing for you. This let alone possible imcompatabilities and other issues. I just will not lie and say the C2Q is always an upgrade, it isn't.

95% of user or even better are way better of with a C2D extreme than a C2Q in mobile form. Some present users with 3.0 GHz C2D's may 99% of the time not relize an upgrade from even a QX9300. and unless they get to 3.2GHz would loose some of the snap...............

 

Well I have an upgradable GPU so I have to think about the future.

The fact is there are many games out there that are at least 3-4 threaded which gets rid of the frequency advantage. A few of those games show a massive increase (games like farcry 2) and its likely this number will keep increasing.

Even if they are optimised for 2 threads the game will get its own 2 cores while windows background gets the other 2.

Windows is very snappy with a quad core, in fact it stays snappier for longer doing more tasks especially when combined with an SSD.

 

I have a SSD so actually for most tasks the C2D at the same speed is snappier. I will agree with heavy multi-tasking the C2Q can be faster. Most users will not load up the CPU to the point needed to see this. Also if you want to compare head to head you would need a comparison of either max stock mobile speeds or attainable overclocked speeds, not a close clock-to clock. I only state clock for clock the C2Q in a single unoptimized thread is less efficient than a C2D.

I am willing to bet those bench's in Far Cry 2 would be closer comparing a C2D at 4.0 GHz compared to a C2Q at 3.2 GHz. Even better a stock C2D @ 3.06 GHz to a C2Q at 2.53 GHz. I know for the power people this is not popular view but I call it as I see it. For 99% of users in the mobile platform C2D x9100 is a better choice than the extreme quad cores.

I am not saying if you have a good running extreme C2Q to change to a C2D, that would have to be a choice a user would have to make. I don't think the return would be worth it in the end. If you have a q9000 though a X9100 might be something to think about..............

 

Some really good info OP. I miss my gateway 7815u like crazy right now. I use to run it at 3.4ghz and did have overheating issues unless i used a cooling pad. Without the cooling pad, after playing for 10-15 minutes it would start to bog down but i didnt remember if i replaced the oem thermal paste or anything when i swapped the CPU.

I went from a 3.4ghz c2d geforce 9800m gts to a quad core 2.3ghz with ati 5650 and my framerates took a huge hit like from 70fps to 50fps. hopefully my new asus will yield some good overclocking results.

 

I don't have a Gateway laptop (HP here), but I chose to upgrade to a C2Q because my 2.66 Ghz C2D was being 100% loaded all the time. I got a great deal for a QX9200, and I've practically never been able to overload it - if it ever goes to 100% at 2.4 GHz, I just up the clock to 2.8 and it's cool .

It's pretty cool, temperature-wise, too - I can run it at 2.4 GHz at 1.0625V and 2.66 GHz at 1.375V - you guys are lucky to have yours run at 1.05V at 2.66 .

I've been experimenting with running the two dies at different speeds - one die overclocks higher than the other at the same voltage, so I can run one at 2.93 GHz and the other at 2.4 GHz - I wonder if that gets me the best of both worlds? Obviously, it runs cooler overall and faster in benchmarks, although I wonder if that affects the stability long-term...

 

You should ask Uncleweb, the creator of ThrottleStop. I am amazed for the simple 1x multiplier you have to jump so high on vCore. The one true issue with cores at different speeds is single thread latency. Where the load is split at 25% on all cores in the scheduler this COULD cause issues. I am not sure if with the different clocks too is the vCore can be set individually. I have not experimented with this as there is no need here.

If you get the AMD Bulldozer patch, there is a 5% increase or so in single threads. It gives a nice little boost in the snap of Windows 7. What we truly need for the core is the true Windows 8 scheduler. I have a thread over in the hardware forum on that. The nice thing too is since the new scheduler is more efficient the CPU can run just a tad bit cooler as well. Also if they do as I've suggested and run the threads dedicated to a silicon core it should eliminate what I said before about latency issues with different speed silicon.

I run 2.93 GHz @ 1.0625 normally now. I need 1.2250 vCore for 3.2 GHz. All of this, even for the short runs before overheating, to do Prime95 without errors. Good news too though is if you use TS windows 8 does not have an issue with it.

Now I wonder when the royalty checks will start rolling in from using my idea on multiple silicon packages such as the C2Q. I'm quite sure Microsoft, and/or Intel, and/or AMD will all look me up just to do the right thing............

 

Thanks alot for sharing you knowledge, Tanware!

 

2.93 GHz at 1.0625V? Must be a QAVR? Damn, I might just consider getting another Q9200 and trying it out ...

 

My dinky i5 will hit 85C when pushed by GPU on my single pipe. I have been thinking of an i7 drop in but the temps scare me a bit.

Check the temps on this guys MSI; NEW Intel Core i7-2720QM (ES) 8M 2.2 GHz (3.3Ghz) Ship Worldwide! | eBay

I've been playing a CPU heavy game that is kicking my i5 down. All the quad guys are getting 20 or so FPS more, however; all my favorite stuff is still not optimized for more than 2 cores so I'm good till that price drops a bit more

 

Yepper, QAVR. Runs well here so far. With the new patch, being as small single thread hits are a bit easier on the CPU, I no runn boot option as 2.93 GHz and maintain under 50C idle. With a full patch who know. I pulled Windows 8 CP from the systembut I still want that full scheduler........

 

Is there a reason you didn't include the t9900 in this comparison? Is the x9100 better?

 

The T9900 is actually a little better stock but more pricey. The x9100 is an extreme cpu with an unlocked multiplier which in the long run is better performance compared to the T9900.

 

The T9900 and x9100 run the same at stock clocks. The problem may be with the x9100 is the steppings available. Not that one will run better but the E0 should run cooler than a C0 stepping.

 

I know that they do the same stock clocks but just overall the t9900 runs a little more efficient with processing benchmarks and games ect of course its hardly a difference that people would notice.

 

In the same machine, not just model but actual machine, the two should do the same other than with the C0 stepping declocking from overheating etc. About the only difference is the T9900 has the Intel Trusted enabled and the x9100 is unlocked. Same transistor count etc. Again so long as the stepping is the same they should perform the same minus heat issues.

Synthetic benchmarks can be deceiving to say the least. Even on the same system the numbers will have a statistical variation and differences you speak of can easiliy fall into that catagory. I was, before the x9100, lloking at doing the T9900 but with the Throttle Stop program the x9100 was a no brainer............

 

so between the T9900 and the X9100, the X9100 is the logical choice?

But if I'm reading this right, it seems the T9900 runs cooler and is more efficient with a lower power draw?

 

Not really. It is not the CPU model but the stepping. The latest, and final release, was stepping E0 for both the T9900 and x9100 as well as the Q9xxx CPU's. Now there are slight variations within the batches of CPU's but generally the C2D's run close to one another as the CQD's do to one another at equal clocks.

With the Extreme CPU's Intel unlocked, supposedly, the best of the runs and spec'd for better TPD's. Of course OC'ng to beyond spec speed will use up more power and more than likely lower efficiency.

In the end at stock clocks the T9900 and x9100 are equal at Stepping E0. When I purchsed the x9100 was more expensive than a T9900 and there was no reason without throttle stop to pay more. That has changed since available x9100's are mostly now at C0.

Off topic Gateway failed with the systems CPU heat sink. While revision 2 works well with the 45w TDP it is designed for the C2D and not CQD. The copper insert design is all wrong. Shims help slightly but we really needed a new HS. Since the P79xx was the end of C2D and CQD life cycle I can see why Gateway did not invest in an entire new design.

Also with the P79xx while they increased power availability to the CPU they lost power to the USB and this has caused issues for some. This can be proven by some older posts I had where it is shown the same x9100 would run stable at much reduced vCore settings.

To compound power issues they used the same 120w brick from the P78xx series. While both graphics chips are rated for 75w with the added power available to the CPU it is easy to see where the brick would be strained to provide enough power for the rest of the system.

So to my solutions;
1.) 180w Targus Brick.
2.) shims on the CPU cores and NB. (ICD as the TIM)
3.) added small copper HS to the heat pipe to offset overheating of the small copper insert in the original HS.

Now these cures may not work for everyone, but they do for me. I run the CQD as default 2.93 GHz @ 1.1625 and power saver enabled. Idle on all cores is well under 50c and normal usage I rarely even see 70c. Gaming max is about 85c, for the limited games I have................

 

^^^ So all things being equal, at stock speeds and with no future desire to OC the CPU, between the T9900 and X9100, which is the overall better chip in terms of performance and efficiency?

 

Neither if both E0, unless you need the Trusted Intel execution then the T9900................

 

So bottom line I should be looking for an X9100 C0?

 

Hey, that's pretty much what I did on my laptop. Do you have a post or thread with more information and details on the mod?

A T9900 has a higher chance to be the better chip IF you don't undervolt, but you shouldn't brush aside overclocking - it helps make an old laptop last longer, and the X9100 is the better choice in that case.

 

No, the C0 is the older stepping. It will run fine and be just as fast but more than likely run hotter. The advantage of the x9100 is you can undervolt with TS but there are apps that will alow you to undervolt a T9900 as well.

Always try for E0, with either chip but as mentioned if you are going to OC the x9100 is your choice unless you are planing some hard mods to the main board.

Edit; some had mentioned with the x9100 C0 only being able to get 3.2 GHz with our systems because of how hot the chip ran...............

 

Ok, thank you.

Will the stepping of E0 be differentiated in the product number? Basically, how can I tell the difference if I'm looking at an E0 or C0?

Thanks.

 

yes and no, CPU-World is one of the best sources of info. as an example for x9100 C0;

QBDW (Intel Core 2 Extreme X9100)

And E0;

http://www.cpu-world.com/sspec/QH/QHBQ.html

Edit; a used QS one at http://www.ebay.com/itm/Intel-Core-...ing-/320892839995?pt=CPUs&hash=item4ab6b4283b

 

That used ebay E0 looks pretty beat up.

 

dont worry about the looks, its just the old thermal paste. If you clean it with isopropanol it would be good as new

 

roger that. Thanks.

 

What is a safe operating temperature for the X9100, both at idle and under load.
I know the TJ Max is around 120*C but I have no intention of ever letting a chip run that hot.

 

The X9100 has a maximum temp of 105 degrees, at which point it will shut down, but anything below 80 degrees is safe for the CPU...

 

Good to know. Thanks man.