Why U500VZ get GDDR5 while N56VZ get DDR3?

does anyone have an explanation for this? just doesnt make sense to me, the U500 is a ultrabook, and N56 is asus' multimedia laptop. a company's ultrabook out performing its multimedia notebook all in the same generation? i just dont see a sense in it. Also i believe asus lost a good number of casual gamers/semi-serious gamers because of that DDR3 on the GT 650m (i for one gave up on asus after 8 years with them and went clevo)

 

*shrug*

Here's as far as I can go on the n56vz chassis (without extra custom cooling ribs on the memory chips, overvolt mods, etc.) before getting artefacts and blips. Meaning it's the limit of the ram and the system's cooling on stock. Max gpu temp 76 degrees:
NVIDIA GeForce GT 650M video card benchmark result - Intel Core i7-3610QM Processor,ASUSTeK COMPUTER INC. N56VZ score: P2802 3DMarks
Score
P2802 3DMarks
Graphics Score
2613

So anyone with an u500 -- go ahead and beat the score..?

And then we can have the discussion again afterwards.

 

doubt the U500 can overclock much with that little cooling, but still the fact exists that its a slap in the face for the people who bought the N56vz now that asus rolls out an ultrabook that can perform better. also keep in mind the 650m with GDDR5 is essentially 660m (with lower clock), and you could overclock this 650m to match the performace of the 660m and maybe the 670m in asus' so called "ultimate gaming machine". Asus' strategy is probably to poach on people without much computer knowledge, and think "ooo 670 much higher than 650, gotta get me some of that!" without knowing 670m is actually old fermi technology

 

If they released the U500VZ right after the N56VZ at a comparable price point I would agree that the N56 buyers got shafted. But it was released much later and at a significantly higher price point. Companies release newer and better products all the time.

 

@silverfern: Except... it's you guys who are insisting on the ddr3 version being inferior -- directly next to a score that is miles above stock as well as safe gddr5 overclock. The 670m also is, without a doubt, faster than the 660. It'll give you some additional 250 3dmarks in that test. But it is also much hotter. So it's not a coincidence that it's only put in the ROG laptops with the dual exhaust cooling, etc.

That's where the kepler cards are so awesome - they're incredibly efficient, and can be scaled down dynamically as well if the loads are smaller. This is incredibly useful in office contexts, as well as with gaming.

But again - u500vz is exactly what "people" have been wishing for, on this forum as well: gddr5 ram, smaller chassis that looks like a retina macbook, and a powerful graphics card.

@sockfish: But - again - beat the score. And then talk about who got shafted. It's right there - that benchmark singles out the gpu score, and lets us disregard pretty much all other factors outside the graphics card component. The score is there, and you will indeed have a hard time beating that without volt-mods and extra cooling.

And yet - you're convinced the ddr3 version is inferior, and that the gddr5 version would have been better. Why? You also know by now that the gddr5 version will generate more heat, and draw more power. So why is this -- even now that we have the stats and the reviews - why is this the better deal?

 

I fully agree with Nipsen.
In fact, all 3Dmarks I've found so far for U500VZ/GDDR5 vs N56VZ/DDR3 show a (surprising?) parity in graphic performance at least (and, yes, that DDR3 P2802 score is at top of my collected results set for the time being).

I'm not qualified enough to declare how much superior is GDDR5 or not in general, but if I could consider this comparison as a significant one (and why not?) I can't see any real-world advantage with GDDR5, at least on GT650M.

Maybe Asus would have made a better choice using DDR3 in U500VZ too, likely getting lower temps with quieter fans!.
But I suppose that a lot of people would have thougth it were a horryfing choice to use "inferior" DDR3 memory!

 

Nipsen, what are you scoring on stock clocks? +135 on afterburner without additional settings is what I scored P27xx with. Not sure why you're making this into a "MY COMPUTER IS BETTER!!" argument, they're just laptops for crying out loud.

Getting back on topic, the S15 scored very high with a bios modded overclock, but so did people with DDR5 640/650/660 models. I would be more concerned about the thermal limit of the CPU rather than the GPU, as the tests for the UX51/U500 show that the GPU maintains a decent temperature and does not throttle while the CPU throttles at 80.

 

those are synthetic benchmarks, if you look at in game performance you will see a large difference. There is a site which compared DDR3 vs GDDR5 of GT 650, the 3dmark11 and other benchmarks showed a 0.5% difference, while in game, it showed up to 20% difference, in some cases, over 20%.

also i wont be criticizm asus so much if these two laptops are from different generations, but the fact is they are from the same generation of intel and nvidia chips. the real difference in price is because the U500 has a 512gb SSD in it, which would cost a bit over $400 (maybe?) when you buy it from sites like newegg, asus would obviously put a good amount of surcharge on that. so if you look at n56 ($1250 at release?) + $500 gives $1700, which is close the price point of the U500 (not sure how much U500 is though, guessing $2000 in north america?)

edit: forgot to mention, the response people got from asus about why putting a DDR3 in N56 was that it was a multimedia laptop, if you wont gaming laptop, go to the G series, and now they put a GDDR5 version in ultrabook, just gives me a whole bunch of question marks and "huh??"

edit2: http://forum.51nb.com/viewthread.php?tid=1216176
forgot to link

also there is NO reason for asus not to put a 680m in the G series (or 7970m, or even SLI/crossfire). the only reason they don't is because they want user to upgrade as often as possible

 

What I said was that if you wanted the coolest and least power-hungry laptop, while still having reasonably good performance, then you would pick the ddr3 version.

It's just annoying when someone insists - while looking at numbers proving them wrong - that the gddr5 variant is better, has higher performance, and is a better pick for the u500 chassis in terms of temperature and heat. It's very obviously not the case: you won't reach better performance even on the stock clocks. And then it still gets you 60+ degrees on the keyboard. It's not a good idea, and it never was.

And if the gddr5 ram version of the card is better performance wise, you should be able to beat that 3dmark11 gpu score, no problem, with a +135/+235 clock. If you can't do that, then what does it mean?

We also know on beforehand that the cards score the exact same score on stock clocks. So are there a huge number of options here? A lot of wiggle-room for alternative theories?

Source? Are we talking this one single unsourced blog again, that put out a few tests between completely different hardware configs and detail settings..?

And the 3dmark11 test is a stress-test, it's not a "synthetic" mark in the usual sense. It runs various routines that all 3d games will use in exactly this way. It also successfully splits the cpu and gpu scores, so we can see how the graphics card performs without having the cpu and ram interfere too much. It's proven to be consistent, and a pretty much perfect indicator of what sort of performance you'll get between the cards. Specially so on dedicated cards. I've never seen one of those results not correspond to fps and detail settings.

And one more thing. Please don't claim out of the blue that gddr5 ram is better on high resolutions without proof. Test it and show me some data. Otherwise... I mean, seriously: "test I've heard about show...". They don't. Tests show exactly what I've given you in this thread. It's right there, you can check it yourself.

edit: In that link you have, they also test the two versions of the card on different hardware - i3 vs i7. And in the 3dmark11 test, the gpu score is 2145 vs 2156. The combined score is miles over because of the i7. But the gpu score is virtually identical, even with that huge spread on the hardware. Remember the i3 also doesn't have the on-die memory controller.

..and no one else could point this out? Or if someone did - the data is just not convincing enough, I suppose?

 

since i dont have the cards i wont be able to test it, but there are people who have as you can see from my link. its obviously not just some made up numbers, since it does show screenshots as proof, the GDDR5 does excel in actual games. also i am not sure who said GDDR5 is better on high resolutions, but i dont think i said that

i also agree the 650m isnt a good choice for U500, maybe asus is doing this for the sake of having the most powerful ultrabook? that seems kind of vain

 

Just making sure. It's usually the next unsourced claim that comes along.

And no, your source does not show that a gddr5 card performs better than a ddr3 card. What it shows is that an i3/650m ddr3 scores a lower total score than an i7/650m gddr5 combo. While it has the same gpu-score in 3dmark11 - and laughably close scores otherwise as well, even with the huge difference in cpu and very likely memory performance..

edit: re. "most powerful notebook".. You do know that several other manufacturers offer a 650m variant in worse cooled chassis, right? Including Apple, just to point that out.

But the idea here that the gddr5 version was:
-faster
-a better choice
-because ddr3 is terrible

Is obviously very wrong. I just wanted to point that out, now that we have the data to prove it conclusively.

 

if DDR3 is on par with GDDR5, then why bother putting GDDR5 on higher end cards? why dont we just put DDR3 on GTX 680 690 to save cost? These big corporations are all about money, if there isnt a reason to invent/popularize GDDR5, why would they do it?

 

Okay, you should realize there are some people on the Internet who like to argue a point to death until everyone is sick of hearing them. GDDR5 is better than DDR3. AMD dumbs the whole concept down so we can understand. Yes, they are trying to sell us more expensive GDDR5 cards, but they are better.

The reason the UX51VZ gets GDDR5 is because you paid twice as much for the laptop. GDDR5 costs more than DDR3. Asus would have increased the price of the N56VZ by about $50 for GDDR5.

 

Intel spent a lot of money on making the Ivy and Sandy bridge on-board graphics cards. And they weren't worth it, and fits into some trail .. far down the line where they finally retire the current industry standard and drop the southbridge/northbridge/memory controller on separate modules.. At which point I hope ARM sues intel out of business.

But where gddr5 ram makes sense is on desktop hardware, and at least cards with broader memory interface buses. On the gk107 chips (640m/650m/660m) the memory interface doesn't seem to be made for that. So you have a memory module with higher internal operating speeds that achieve the same throughput, except it does so at higher temps and higher power-draw. But on a desktop card or a different chipset, having gddr5 ram might make sense, since 10 degrees doesn't mean anything, and the volt can easily be set higher, the setup can use a broader bus and much higher clock-speeds, etc.

The best reason to use gddr5 ram is probably cost(edit: manufacturing cost is lower). To strip off some of the unused features ddr3 ram has on graphics cards in general, like software controlled volt, low latency options (graphics cards won't benefit from that - not how they're designed), that sort of thing.. So it makes a lot of sense to manufacture less expensive ram with higher throughput options (and higher latency) at higher clockspeeds.

But on a laptop, it might not make sense over ddr3 ram on that chipset. On a different graphics card chip, you might get better performance. But you still get higher power-draw and more heat..

 

if you look at the 660m, nvidia advertise it as a gaming card, same bandwidth at 128bit, but you see no DDR3 option, again, why doesnt nvidia slap on DDR3 and off it goes? Also GDDR5 has higher frequency than DDR3.

I dont think the onboard video from intel is uselss, the HD4000 works really well when on battery. if the dedicated card is used the battery would die too fast

edit: found this review, between a 512MB GDDR5 vs 1GB DDR3:
http://www.hardwaresecrets.com/arti...B-GDDR5-vs-1-GB-DDR3-Video-Card-Review/1272/1

GDDR5 version, even with only half the memory performed about 10% better more modern games (crysis warhead), performed better at higher resolution (note, even with less memory)

 

They also rolled out a 640le version with the same chip (the gk107 chip), with a bios locked to lower frequencies. This one is possible to overclock until it reaches pretty much the same scores as my 650m. They didn't advertise that, of course. But it's how they roll out products. Having several different variants of the card so they can cover more demographics. And if it costs less to develop the same chip and just change the bios on it - well, then that's what you're going to do.

Same as intel has done with their processors for a very long time. Back in the day, we had pentium 90, 110, 150, 180, and 200..? I think. That was the same processor where the only difference was changes in production process. Exactly the same blueprint, same yields, same everything. The celeron versions had some features physically locked.

The reason for that was to cover a larger number of potential customers, while milking the ones prepared to pay the most for the best performance, of course. It makes perfect business-sense, but it doesn't make the actual product better for us as customers. So I'm just saying that "they made it, so it has to be better", is a good argument.

But yes, gddr5 memory has potentially larger speeds and higher throughput on a properly configured memory bus. It does appear, however, that this is not the case on the gk107 chip. And the reason for that very likely is that the gk107 chip is not made with different memory bus interfaces. But instead is manufactured with one single interface for all the variants.

Does that make nvidia frauds? Yes. But whatever. If people were informed customers, they'd pick the best products anyway. I mean, be aware of that this isn't industry secrets. The gddr5 whitepapers are easy to get hold of, and the uniquely good oc results for the gk107 chips are widely known.

Actually, not on the kepler cards. If you measure it, the same task on the kepler card with the dynamic underclocking actually draws less power than the intel graphics. On the 670m and the fermi cards, since they don't have the dynamic underclocking, the intel graphics are going to be useful, though. Performance, shader standard implementation, feature-wise, etc., etc., it's a terrible card, however. I use the 650m for video-acceleration while on battery because of both those things (also, it can't actually be turned off, so why not).

Well, this is a different chip again. They also don't have identical hardware, apparently, and they do test games that depend on memory and cpu performance. And it would have been interesting to see the gpu score in 3dmark11... because these scores, again, are just too similar.

I mean, if the bus-width was doubled - which is what potentially you would get with gddr5 ram... then would you expect 10% increase? Doesn't make sense.

 

not sure what you mean by dynamic underclocking, but when i am on battery (or idle), my 675m goes to 51mhz core and 135mhz memory.

also i didnt read the whole article, just results, but i was under the impression the hardware are identical apart from the memory? you have to keep in mind the memory size is halved, so for more demanding games like Crysis, the GDDR5 version might not have enough memory to show its full potential

 

The kepler thing is about dynamically clocking up and down depending on load. Say you run a program that needs only some of the processing units to reach 30fps. An old game or a 3d editor, that sort of thing. Then the gpu is for example only 10% active, with a lower clock, and lower power-draw.. even when it's "powered up". Very neat, imo.

I know I'm contradicting absolutely everything you say now. But if anything having less ram would mean perhaps less pre-loaded assets, and no redundancy on the memory transport between the physical dimms. Those on their own could actually increase the performance. But generally ram size has nothing to do with it.

What I'm wondering about is how a ram type that on paper has twice the bandwidth, can somehow never seem to get more than 10% increase in performance.. it doesn't seem very likely we're really seeing the difference between the ram. And more likely it's something else.