[Exclusively for NBR] NVIDIA Kepler GK107 and Intel Ivy Bridge BENCHMARKS revealed by yknyong1!

It is a paper launch. You are only buying it March 26 at the earliest.

 

P1719, told ya guys

 

When Alienware M17X R3 first came out, it was offered with 2XGTX460M in SLI or a single 6970M. Then later it was offered with 2xGTX560M in SLI or a single GTX 580M

Alienware M17x R4 will have two configurations. Either 2xGTX675M in SLI or a single GTX 660M.

Thoughts?

 

oh yes I know, and that is 680 (not mobile) mobile chips will sure take longer...

dude that is wrong, r3 never had SLi or xfx, it is single GPU system... (enough though)

 

oh DAMMIT. And I thought I was on to something here

 

GTX 480 was released March 26th 2010, GTX 480M in May 25th 2010
GTX 580 was released November 9th 2010, 580M in June 22nd 2011

New architecture (Fermi) came with the 400 series, so lets hope it only takes a couple of month at max to get the mobile version out this time too

 

there's no pattern, and you shouldn't even make sense of that, it's pointless, I don't think these companies follow any pattern.

 

I didn`t say it was either. I was just looking at the dates, hoping (thats why I wrote "hope" above) it doesn`t take any longer than the 400 series. I`d hate to wait so long as the 500 series took

It is either GTX 660M, 680M or 7970M for me. If there will be any mobile GCN from AMD this year
I`m in no hurry though since I will wait for Ivy Bridge anyways

 

well they had the 485m in between too

 

Ivy bridge will supposedly be out April- Early May, and the ETA on the new Clevo models is April 12, but I'm sure Clevo knows a lot more than we do so maybe it will have at least the new Intel CPU + 600 fermi since I realllllly don't see why Clevo would release a new model without it actually being the new gen.

 

Maybe it's just a preorder on April 12.

 

People, game for Cinebench?

 

How many benchmarks did you manage to get at the fair? The fair already ended yesterday

 

I am not Superman, how many did you expect?

 

The way you've teased this community is just awful. It seems like you have a very high demand for attention, and seem to focus more on yourself rather than sharing the information you have.

By all means, kudos to you for getting these benchmarks, but the way you chose to publish them just leaves a general bad impression.

You certainly won't be getting any positive reputation points from me.

/rant

 

If I really wanted to tease people, I would not share my 16? pictures of specs at one go on 8 March. Additionally, I wouldn't have released results of three different benchmarks within a day.

Benchmarks were done on 10 March. Immediate release of 3DMark06 when I got back home from the crowded fair. 10 to 11 March. The sad news is I lost the Vantage picture. I could have easily spread everything over two whole weeks if I'm the attention seeker you are thinking of.

Fritz Chess no one was interested, so for 3DMark11 I tried to make it interesting by having the fellows guess the score. But it failed badly. Red was trying so hard, so I just released the benchmark already.

Guess I can't really have everyone happy.

 

You actually had the Vantage score?
CURSES!!

 

thanks for the benchmarks bro.

 

Thanks for your time man, I don't think a little playing game can hurt anything, after all this is not a service yknyong1 is providing to us, we should be grateful that he took the time, avoided acer personnel and brought these pictures to us, all appreciated by me at the least! (repped already )

 

Thanks. Repped you for expressing your gratitude with sincerity!

 

It was in the mid 8000s IIRC

 

So taking what we've learned here, in summary, Nvidia more than doubled its mid-range performance, in replacing the GT 540M with the GT 640M.

It's an ugly, brute force method, but you can't deny the results. I'm seriously impressed by what the move to 28nm has allowed them to do. No one thought the 540M's successor would have as many cores as the GTX 580M.

This makes me hope the GTX 660M is more than a 640M with GDDR5. Give us 448 shaders and decent clocks, probably 192-bit GDDR5 would do.

 

I'm still not convinced that GPU-z is reading the number of cores correctly. Even with the die shrink, quadrupling the number of cores seems kind of implausible.

 

The Nvidia control panel surely reads that correctly. Just in case, I have already checked the retail version of Acer M3 and it shows 384.

 

Kepler processors are different than Fermi processors apparently, that is the consensus

 

Core for core, Kepler cores are weaker than Fermi cores, just like how Fermi cores were weaker than G92 cores.

This also means more cores for NVIDIA to achieve parallelism, essential for computing.

 

are you certain about this? I have been gaming exclusively at 1080p with my AMD 5830m for the past few years and the NVIDIA 550m is rated higher according to notebookcheck

With the exception of Witcher 2, I have been able to run every game I've tried at that resolution with decent framerates, sometimes with FXAA instead of MSAA (or reduced AA) but often with all settings completely maxed.

 

Just hope that it is GK104. I've heard of GK106, was that cancelled?

 

If the fermi cores are more efficient clock for clock, why didn't they just do a simple die shrink on fermi to get the power down so they could booth the number of units?

 

Lets hope this brute force comes not of the cost of heat

 

Well the 640M is already in a few less than 1" Ultrabooks, which is unprecedented as far as dedicated graphics go, so I have to assume that the heat output is pretty low.

 

It doesn't even feel warm after I ran so many things on it. Makes me wonder if the Turbo or hot clocks can be varied by the manufacturer.

 

GT 640M preview (benchmarks mostly):

Nvidia GeForce GT 640M Review: Kepler Arrives For Mobile | PC Perspective

 

According to this article the Acer laptop gets hot as hell.

 

Given the shear amount of cores and the tremendous gflop output packed in this beast, its such a waste of computing power that nvidia decided to neck the bandwidth by using a pathetic jurrasic era ddr3+128bit combi.

One could only imagine it trashing the 560m and matching the 570m if packed with something more decent and 21st century such as gddr5 or a 192bit bus

 

I bet NVIDIA did this to save money. Using GK107 (which is already the lowest performing core for Kepler) would save the need for a new lower performing core. Using DDR3 is to brute-force performance at the intended level. (GT 640m or GT 650M)

 

That is the Timeline Ultra M3, with a thin-profile ultrabook like this, the tradeoff would be heat.

The one tested by me is a normal 14" notebook. Acer Aspire V3.

 

Like I said, thats extremely hot! And they arent even talking about the GPU itselve but rather about the exterior of the notebook like the palmrest, the bottom and so on.. In combination with ivy bridge it probably throttles too!

The 555m in my previous m14x never climbed over 65 degrees

The few benchmarks you run dont really tell if it gets hot or not since it was only for a short time (under 10 minutes), you shouldve run prime 95 and furmark at the same time.. I suspect the similiar heat for this notebook

This review for example tested the same notebook and claimed the notebook was cool, but only because they didnt do longrun tests!

Laptop.bg Review

Remember the CPU it uses is a ULV i5-2467m! So the CPU should be relatively cool at 17W TDP!!

 

One more GT 640M review at Anandtech

Intel HM77 spotted!

AnandTech - Acer Aspire TimelineU M3: Life on the Kepler Verge

"Acer actually broke the embargo and began selling the notebook early, causing the rest of the press (and NVIDIA) to scramble to put together these reviews."

Kudos to Acer, which enabled the leak from V3 BTW. (Repercussion unknown)

 

Also, from the review, it is said that the top cover does not heat up much. A "sharp heat" was described, but not sure what that means.

 

Benchmark 4 - Cinebench R11.5

 

Nice. That's like 10% faster than a 2630qm.

 

Benchmark 4 - Cinebench R11.5
for Acer Aspire V3
Intel Core i7-3612QM 35W 2.10GHz 22nm Ivy Bridge
NVIDIA Geforce GT 640M 2GB DDR3 GK107 Kepler
Default Settings for all.

Scores with GPU comparison


Scores with CPU


As we can see, Ivy Bridge provides only an incremental jump in performance over Sandy Bridge.

However, the GT 640M is actually almost 30-50% faster than its predecessor, the GT 540M.

This is the end of all benchmarks, hope you all enjoyed yourself having a first hand look at newly-released hardware!

I would appreciate if you could provide me your insights so I can do a small writeup for the first post.

 

Wow! 35W 3612QM is faster than 45W 2820qm in CB 11.5.

And the 45W version i7-3610qm got 6.09 in CB 11.5, higher than 2920xm and fx-8150! Poor AMD.....

Intel Core i7-3610QM Tested | techPowerUp
Can't imagine the performance of 3720qm/3820qm/3920xm.