Projected R3 versus R2 bechmarks

IMO, if you don't want to deal with all the overclocking silliness, then you should just get an R3 and you'll be running games perfectly fine at stock speeds. A good example is Starcraft 2, us R2ers are required to OC if we want to play 4v4s, whereas those with an R3 will be able game right out of the box.

 

Ok guys Anand has recently tested another Dell laptop with Sandy Bridge (albeit quad core) and 540M

Check out this link for a flavor of what the 540M can do:

Dell XPS 15 L502x: Now with Sandy Bridge - AnandTech :: Your Source for Hardware Analysis and News

I would say to subtract probably 10% or more because of the faster processor in the XPS L502. But yeah the benchmarks are mostly at M11x resolutions and lots of games are tested with lots of other notebooks with similar GPUs. This ought to give you a feel for whether the speculation thus far about the R3 performance is right or wrong....

 

So your dead line is 26/4?

Would you like to see the full picture?

The user reviews won't make it before your deadline. Go order it if you have no choice.

The overclocking is really for the established R2 users to get extra speed. You're not required to OC to play games though. Installing Throttlestop is the most basic step to improving performance.

 

My WoW/SC2/BFBC2/L4D2/DOW2 installations disagree with you.

 

For any resource heavy game, you need to oc the r2, plain and simple. Unless you find pleasure in lag? Throttlestop isn't going to turn a 2.2ghz into a 3ghz.

 

Yes. This.

The problem is that people aren't measuring the dual core speed correctly and just counting the single core max Turbo as the clockspeed when it's not just because it's bigger. At stock, with Throttlestop, a dual core load will have an i7 running at 1.86Ghz (14 * 133). With the max 166 overclock it's running at 2.33ghz (14 * 166). While 166 is a 25% overclock, it is a far cry from night and day difference- The stock i7 can still play anything the OC'd i7 can.
Throttlestop is not going to turn 2.2Ghz to 3Ghz because, in games, the CPU won't be running at 2.2Ghz to begin with unless the game is ancient and single threaded.

 

True for the R2 but sandy bridge has a whack more IPC.

 

No real deadline, was just hoping to order by the end of next week. It's just the patience issue.
I'm not so sure about this full picture you mention
You always have a choice - even when the options are limited - so I'll wait for a couple more reviews.

I'm most interested in battery life still ... and I'd really love to see a review of a release model, not all these pre-production models. Anyone else? Maybe there's one out there I haven't found?

 

You should look to the past and see earlier UM CPUs throttling under long loads.

Smart move. You're the only person I've seen that recognised that the review models are non-existent and not indicative of shipping models.

 

M11xR1 at 2.0Ghz when the base clock is 1.3Ghz. Over ⅓ increase on standard.... not to mention the deserving golf clap from your pairs on completion, which can never be overrated

"And while you’re out, play up to 10 hours² on a single battery charge." - As stated on Dell web page.

10 hours battery life in a pseudo quasi environment known only to one boffin in an unvisited office, seems so bizarre it wouldn't surprise me if the M11xR3 could manage some impressive battery stints.

One has to remember they only had 8 hours listed for the M11xR1 when it first came out. So at least, can we read into this that the battery life is as good at the M11xR1.

² - Based on testing using the Mobile Mark 2007 battery life benchmark test. For more information about this benchmark test, visit www.bapco.com. Test results should be used only to compare one product with another and are not a guarantee you will experience the same battery life. Battery life may be significantly less than the test results and varies depending on your product’s configuration, software, usage, operating conditions, power management settings and other factors. Maximum battery life will decrease with time and use.

 

So what I gather from this thread, if I have and R2 i7, I should be more than happy with it and not be disappointed that the R3 came out only 4 months after I purchased it?

(Bearing in mind that I grabbed the R2 for $1550 AU, after a $150 cash back in January of this year (yeah, I know it's so damn expensive here), and the best model i7 R3 is up at $1800).

Just for reassurance purposes so I can sleep soundly at night Haha, thanks.

 

Yes. Also, to get the most out of that i7, make sure you OC in the BIOS and install throttlestop. You may not reach the heights of the R3 in performance, but you'll close that gap considerably.

 

Anyone know what the expected battery life of a r3?

 

Well to be fair, won't that change once the R3 is out in the wild? All the tools/methods out now for overclocking the R1/R2 didn't exactly roll out the first month of the M11x release. There will be a way to get past that...

 

if the clock gen is on the CPU its self... then probabily not, unless you can mod a Bios.

 

CPU-Z shows both my 2 physical cores clocking around 2.2Ghz stock. I believe it's when all 4 threads are being maxed out is when you see a max of 1.8Ghz. -_-

 

Ah, thank you very much. I've heard of this throttlestop but I've been quite reluctant to try it out considering I was suffering from the infamous lockups (you'll see a 33 page sticky about it). However I do believe this problem was specific to the graphics card, and I understand throttle stop is simply there for the CPU, in which case shouldn't be a problem.

Anyways, I haven't really played anything that requires any more oomph, (except ArmA 2, but I believe I'd require more than a little oomph to run reasonably well in all areas, so I'll leave that to my desktop). I'll definitely overclock when something I'm desperate to play on the M11x comes out, so thanks for the advice, I know where to go for tutorials etc on the subject.

 

Depends where and if R3 is throttling. The numbers can't be right for Cinebench 11.5- it's only 10% faster then my i5. What else is 10% faster then my i5? A SetPLL'd i7.

It's hardware locked. Not even Throttlestop can help it unthrottle.

Modding the BIOS won't change the fact that one clock generator controls everything. It will still be bad at overclocking but at least you got the BIOS options like those M17x R3 owners.

The single multiplier only engages at single thread loads. You don't need a full quad thread load to see the dual core multiplier in use. Here, Prime95 shows that with my i5.

Prime95 1 thread: 14x (single core multiplier)


Prime95 2 threads: 12x (dual core multiplier)


Here is Throttlestop confirming it is 14x/12x.


-_-

 

You do realize that even though you created 2 worker threads in wprime, those workers are being loaded on all 4 threads as indicated by throttlestop? You basically proved yes 4 threads do indeed down clock on max load which is expected, yet you haven't proved anything otherwise. If you can, try setting affinity to run on 1 thread for each core and see the results.

 

K done.


Can't help the unloaded thread usage- that's idling.

 

Didn't stop you form pin-moding the C2D. Not saying that it IS possible, but just throwing that out there.

 

My pinmod was possible from the beginning as Intel couldn't control how a PLL works. This... This has been made impossible deliberately by Intel using one clock generator for every clock. We can mod it but we won't get far before something is unstable.

I'd be more concerned with the throttling that appears to be happening TBH.

 

It makes sense, if everything is controlled by one clock, mosdifying that will mess with alot of things. I wonder if being able to completely disable the IGP could make it more stable to OC? Just a possibility, if the discrete GPU dosen't go though the IGP like before. Also, we dont know if there is any possibilities of a hardmod yet, no one has sytems yet.

 

You can't disable the IGP because Optimus is still used. We need the datasheet for the PLL before we can do anything. Any increase in clockspeed to BCLK will increase USB, SATA, PCIE and other clocks. We need something like a FSLx mod to hijack the BCLK signal only to get anywhere.

 

Like on my M9750. I can use Clock Gen to OC, but I can only get to 2.3GHz before anything messes up besause it changes everything, like the MXM slot, USB and what not.

 

Difference is that you can do an FSLx mod and overclock. Not sure about it touching other clocks. A hard mod won't help the R3 users. At least not from what the R3 is experiencing. It will overclock. In kHz. The Hardware Heaven review has overclocked results, go check it out.

 

I saw those. VERY small OC and difference. Hah, in KHz reminds me of the 80's.

 

The "Max supported frequency" in this link should tell you everything about how overclockable R3 is. Not even the 5% I predicted or even a 100mhz gain in a dual core load.

 

Holy... ouch. Well, so much for OC'ing the CPU the normal way. I doubt there will be any (easy) ways to OC these CPU's. Hopefully the GPU is very OC'able stock.

 

And that's the good news.

The bad news is that the R3 is definitely throttling- there is no way the 2657M can be 10% faster then my i5 (2.13Ghz) considering the 20% IPC gain and higher clockspeeds. A good i7 R2 166 can mop it up.

 

I noticed that. It seemed very odd. Did they wreck the cooling in the thing, or is the BIOS bad? Maybe they just need to fix fan settings. I would expect the 2657M to kill the i7, with its better performance per clock.

 

They probably used a different fan table aiming for quietness. I hope. I can't see the heatsink so I can't comment on it. Maybe that's why Dell given out 2657M R3 instead of 2617M? To bias the reviews? Regarding the BIOS, Unclewebb is an expert on throttling so he can explain what can (and can't) be done to fix this.

 

Hopefully he'll post here. I know what throttleing is, I believe, the CPU or other parts are downclocked due to excessive temperatures? I had that happen in my M9750 when I (foolishly) put a 280M in it. It worked but ran at much slower clocks due to insufficient cooling until I did a copper mod. I wondered, also, why they would send a CPU not designed for it.

 

Throttling is exactly as you described except in this case, it's due to power as well as heat. The R2 had this problem losing, all Turboboost at long load, leaving it running at 8x133 or 9x133 before Throttlestop fixed it. Shame it won't happen again. The 2657M is compatible with the R3 and is the fastest Sandy Bridge ULV CPU except Dell is not shipping it. If performance scales, it will be ~4% slower then reviewed. Highly doubt it with all the throttling.

 

Designed, I meant not shipping. I think that it may just be down to fan management, probobly just an under agressive system. I sure hope it can be remedied, since I'd hate to see the R3 screwed over. I think its not really right to have an R2 I7 with TS at 166MHz perfrom better than the new SB CPU's, and definitely makes it less attractive to get an R3.

 

The R3 can support the 2657M. Dell choosing the SB ULV is the first step to getting screwed over. Not much we can do abut throttling, according to Unclewebb.

 

I know. I hope they ship it, but it seems pointless when it runs too hot. And, that sucks.

 

I don't see any way to get around the limitations that Intel has built into their second generation Core i processors. Desktop Sandy Bridge CPUs run great because they have unlocked multipliers but the Ultra Low Voltage mobile processors in the Alienware M11x-R3 don't have that feature.

Intel redesigned Turbo Boost so the maximum turbo power limits are fixed within the CPU and can't be adjusted higher like you could do in the R2. Bus speed overclocking is also limited. If you can't increase the multiplier and you can't change the base clock speed any meaningful amount then overclocking these new Ultra Low Voltage CPUs is officially dead. Bios overclocking is limited to 3% because the current Sandy Bridge chipset generation becomes unstable when you try to go beyond that. No more SetFSB or SetPLL overclocking is possible.

As DavyGT pointed out, the Core i7-2617 in the R3 might not be able to maintain full turbo boost and run at 2300 MHz when fully loaded due to turbo throttling. We'll have to wait for a proper review to see some data on this. The Notebook Check review mentioned zero turbo boost while synthetic testing at full load due to power or heat related throttling or both. It continued throttling down to 900 MHz but I haven't heard anyone mention this yet. Throttling issues were close to non-existent in a properly setup R2, even when fully overclocked.

 

Ok... so where were we?

Ok yea, also, try playing SC2 4v4 non-OC then OC it to 166. It's a night and day difference. Anyone who's serious in gaming needs to OC the r2.

This testing image is better at representing what we were talking about than your previous testing. These results make sense since turbo boosting takes into account active cores and thermal limits, etc. So I'm not surprised to see these results, in fact, it's probably reducing the turbo steps because there's 2 active cores. Actually, I'll assume other variables are appropriately assigned, because I also noticed you're overclocking to 175, when we were talking about stock, but I digress.

However, I'd additionally would like to see how turbo boost is effected when given a low steady multi-process (each with 1 thread) load, or low multi-thread load, and see if that has an effect on the results. Would be interesting to see if the load amount has an effect on turboing.

It should be also noted that it appears throttlestop takes the average multiplier? Which is why I was actually seeing some discrepancy between cpu-z and throttlestop. For instance, when throttlestop shows a 14x multiplier, cpu-z can show a steady 17x multiplier in its cores. Likewise, throttlestop shows a 16x multipler, and cpu-z sometimes shows a 14x multiplier in the cores.

 

If anyone can reach 166, please read my SetPLL guide to OC even further.

That's a different thread in itself.

 

I gave 64bit cinebench a quick and dirty go with my i7 @ 175. Notebookcheck R3 results in parentheses.

CB10-single: 3831 (4376)
CB10-multi: 7644 (7715)
CB11.5-CPU: 2.02 (1.91)

 

NBR doesn't have a review on the R3. Do you mean NBC? Gets close to matching in CB10 multithreaded and beats it in CB11.5. (The score is 1.91). By quick and dirty, do you mean an unoptimized Windows install? Good results nonetheless.
Remember, this is a review unit with a faster then shipping CPU (2657M) so user results will be lower with the 2617M.

 

Oops, I meant notebookcheck. Fixed. And yes, it's unoptimized. I just installed it and ran it, still had firefox, AV, etc running in the background.

 

Ok. You better go and optimize it then. Who knows what kind of gains you can have. Thanks, you're the first i7 TS SetPLL user to step up.

 

Now optimized. Just to reiterate, this is an R2 i7 640UM @ 175 (w/throttlestop at 4/100/100). Notebookcheck review numbers in parentheses. Worth mentioning that the other major difference in our 2 configs is an SSD in the review sample, I have a 7200rpm drive (don't know if that would make a difference though).

CB10-single: 3974 (4376)
CB10-multi: 7770 (7715)
CB11.5-CPU: 2.04 (1.91)

I'd like to see what new R3 owners get with the production CPUs. But this looks very interesting indeed.

 

Very good. You beaten the fastest ULV, even if it's overclocked (Hardware Heaven posted an overclocked score of CB11.5: 1.96).
Any gain between you and R3 will be widened with the production model. Any loss between you and the R3 will be closed or minute with the production model. An even better benchmark would be WPrime 1024 as it runs longer. We can only wait for the user reviews to surface.

A SSD won't make a difference in these CPU benchmarks.

*Cue DrGoodvibes bursting in with surprise and wonder*

 

Those are close scores. The GPU is going to be the major difference, I believe, even if it isn't much different. I don't see the faster RAM making much of a difference.

 

That i7 will pull ahead when the production models arrive. Not bad for the R2 so far- the i5 is ~7% behind and the i7 is ~7% ahead at 175BCLK. It's going to be smaller for the R2 and larger for the i7 by a couple of %. Linear scaling!

It's a new architecture so they're not comparable directly. Besides the reviews, we don't know for sure. It's like trying to compare the GT280 (335 is distantly related to this) to the GTX 460 (540 is related to this) and saying the 460 will be ahead significantly because it has higher clocks and more shaders when it really is slightly ahead, if that.

Correct. There is much greater performance gain from overclocking the CPU then overclocking the RAM. It only serves as a hindrance for the R2.

 

I find it odd that the 460M is only slightly ahead of the 280M, when it ahs less shaders and what not. Off topic, anyways, yeah, they areb't compareable. Its almost like trying to compare AMD GPU's to NVIDIA, you can't without benchmarks and gaming tests.

 

I just updated the first post within this thread.

My update points to relevant posts here that , while does not proving Crysis 2 runs faster with the R3, it does prove that the R2 has enough power to spare to match or beat the R3, CPU for CPU.