NVIDIA's Optimus uncovered in Sony VAIO Z, along with TRIM support?

Sony's illustrious VAIO Z may have a so-called "Dynamic Hybrid Graphics System," but NVIDIA's Optimus it is not. Or is it. Notebook Review's own igorstef started digging deeper within the laptop's switchable GPU setup, and lo and behold, it sure looks as if Optimus is underneath. He went through the painstaking process of installing a slew of new drivers and tweaking some code within select .inf files, and in the end he seemingly found a way to get Optimus drivers working on the new rig. Of course, the debate has been raging on for five pages now, and it still seems inconclusive as to what's really going on behind the scenes. In related news, ZoinksS2k seems to have discovered a way to get Windows 7's TRIM feature working on his SSD-equipped VAIO Z, and if you're interested in doing some tinkering of your own in either case, you know where to dive in.

[Thanks, Bill]

[boom roasted Posted Mar 23rd 2010 9:20PM]

im guessing optimus technology will be on the new macbook pro's

[buri100 Posted Mar 23rd 2010 9:26PM]

@boom roasted

I'm guessing optimus technology will be on the new Dell studios.

[boom roasted Posted Mar 23rd 2010 11:07PM]

@buri100
that better happen haha and the studio xps's. when do you think they will refresh those? i plan on buying a studio xps before august.

[Leindurstit Posted Mar 23rd 2010 9:21PM]

As I noted in that Notebookreview thread, it's not without its bug, namely the lack of the 330m actually being able to turn off. This means that there is literally no benefit over just running the 330m all the time, until further notice.

[samin100 Posted Mar 23rd 2010 9:21PM]

vaio's...

i remember those

[gordonyz Posted Mar 23rd 2010 9:28PM]

Will the i5 ENVY 15 have the switching mechanism?

[bakkermaarten007 Posted Mar 24th 2010 4:16AM]

@gordonyz You do know that the switching mechanism is not a new feature, do you? The previous version of the Vaio Z had it too

[Peytral Posted Mar 23rd 2010 9:52PM]

the vaio Z is the sexiest laptop ever

[(Unverified) Posted Mar 23rd 2010 9:58PM]

Peytral is the sexiest poster ever.

[XRX Posted Mar 24th 2010 4:17AM]

@Peytral

you r the best guy ever.........

[bakkermaarten007 Posted Mar 24th 2010 4:17AM]

@Peytral Let's talk about sex

[(Unverified) Posted Mar 24th 2010 5:32PM]

@XRX

Lulz at the fact that you're probably right.

[spin cycle Posted Mar 23rd 2010 9:57PM]

The poster make the error of assuming that TRIM is the only way for an SSD to become fast again. But it's not the case. The SSD can "trim" itself in the background. And by leaving his computer on for a while after exercising the drive, he let it do so.

TRIM is vastly overrated. A well programmed SSD will see only a small advantage from an OS implementing TRIM. People just are used to poorly programmed SSDs.

[norp Posted Mar 23rd 2010 11:06PM]

@spin cycle That is not true at all.

Trim allows the SSD to mark parts of blocks as unused as the filesystem marks them as unused, if the SSD has no knowledge of the filesystem sitting on top of it it has no idea if a sector has data on it.

Now you can perform this in the backgroud using a bit of software running inside your OS that is aware of how the filesystem actually works but this has absolutely nothing to do with how the SSD's controller is designed.

I could go into a lot of detail but i'm sure wikipedia has a good explaination of TRIM that will refute your claim. Remember that not every harddrive in the world is formatted in NTFS.

[spin cycle Posted Mar 23rd 2010 11:56PM]

@spin cycle norp, you can quote wikipedia all day and you're still wrong.

The problem which TRIM solves is this:

You go to write a page (perhaps 4K) to an SSD. This page is smaller than the block size (typically 128KB or more) of the NAND (including all stripes). But on NAND, you have to erase a page at a time. So in order to write, you need to erase a block, then you have to write in the pages that are written, and then as you wrote more pages in that block those are written in also, and if you never write all the pages in the block, then some pages are in the old block, some are in the new block, and now you have two blocks in use to hold the data of one. Since your device only contains a specific number of blocks, this is clearly a problem. Eventually, if you end up with enough of these partially completed blocks around, when you go to get a new block to do this process, you don't have any spares, so you have to clean one, by merging the two blocks ("old" and "new") that are storing the data that really only needs one. Then the "old" block becomes available for you to use.

But the problem is, that this process of merging takes time, so a program goes to write, and instead of a write, the SSD goes on this big process of merging two blocks before the write can even start. That's the slowdown that occurs over time.

How TRIM solves it is that you can tell the SSD that some blocks aren't in use. Then those become available for picking up to be used as the "new" block in an unmerged pair, so the writes you make go write into those unused blocks and your writes never have to wait on a merge.

However, TRIM isn't magic. If your device didn't run out of available blocks to use as the "new" one, then you wouldn't hit the slowdown case. So another alternative is that your SSD, while it isn't busy doing anything else, can go through and merge "old" and "new" blocks in the background. Each time it does this, there becomes another clean block ready to use for a write. If the device has enough spare blocks and you give it time to "rest" between big chunks of writes, it can free up blocks to use for writes before the next big write. That's what a good SSD does, and that's what this SSD is doing here.

If an SSD is designed well, and it gets a little time to rest between big bouts of writes, then TRIM will produce little to no speedup because you'll always have spare blocks on hand and not hit the slowdown case.

Is that the kind of detail you were thinking of going into?

[Hung Posted Mar 24th 2010 4:46AM]

@spin cycle

I don't think that's fair to say of the poster. Your comment also applies to the tester, who figured TRIM was significant enough to invest hours into getting it working.

But I will agree with you on the point, regardless. It's definitely unnecessary considering these are down pretty good. Either the MLC or SLC (likely the former) is great long-term on Samsung's controllers, making TRIM a nicety, but a nice garbage collector yields the same effect.

[DoctarPeppar Posted Mar 24th 2010 10:09AM]

@spin cycle
Tell that to all the Intel X-25 g1 owners on the Intel forums who complain that their drives slow down after 6 months, and they have to do a backup \ zero out \ restore to get them working fast again.

TRIM IS significant; I don't care what you say.

[spin cycle Posted Mar 25th 2010 12:05AM]

@DoctarPeppar
I said on a well programmed SSD, TRIM isn't important. The Intel G1s are not well programmed. The solution is better programming on the SSDs, TRIM is far less important.

Get a well programmed SSD (I have three!) and you won't care whether you have TRIM support or not. Which will be very helpful since neither Windows nor Mac OS support TRIM.

You'll see well-programmed SSDs roll out widely over the next year.

[trueislander Posted Mar 23rd 2010 10:32PM]

Looking forward to testing the "Optimus" feature out when this beast gets delivered on Thursday! In my opinion I'm not one of those who feels inconvenienced by having to throw a switch in order to turn off/on the discrete graphics.

[fl0ppy Posted Mar 23rd 2010 11:41PM]

@trueislander

I'm ZoinksS2k.

We don't have a working solution for Optimus yet. The slow boot problem needs to be solved to make it practical.

[hdawggy Posted Mar 23rd 2010 11:35PM]

Damn i creamed my pants again... how can i still drop the same load... lmfao after the 12-core AMD... omfg