Is Intel Turbo Memory worth anything?

[barrywohl]

I'm biding my time planning to order a T61p with a WUXGA 15.4" screen, 3 gb ram, Vista 32 bit (business or ultimate) and a 100 gb or 200 gb Hitachi 7200 rpm Travelstar.

I've now read two reviews of Intel Turbo Memory which question any performance improvement. It seems like 3 gb of ram is the optimum for 32 bit Vista and that the Intel Turbo Memory only really shines in memory starved systems.

I almost wonder if my former Z61p with a T76000 2.33 GHz Core 2 Duo processor, 2 gb ram, one 7200 rpm Windows XP drive and one 7200 Vista Business drive wasn't really just as good as the new Santa Rosa systems. I guess in the long run the T61p ought to be a better product than the Z61p. The T61p has the screen roll cage, newer Santa Rosa processor and a bit less weight and bulk compared with the Z61p. It also has fewer watt-hours in each battery option and thus less battery run time than the Z61p. Lenovo has taken back my Z61p for credit due to a freeze-up problem. Well, I'm hoping the T61 planar board is a better product than the Z61m planar boards.

[cshbc]

Any one have reveiews of this option?

[Ivan Ivankovic]

Any one have reveiews of this option?

Anandtech reviewed T61. search it out.

[Tony Chan]

Wonder if it works if we swap turbo memory into a T60.

Anyone knows if caching logic is built into the OS ( i.e. Vista ) ??

Tony

[ryengineer]

Wonder if it works if we swap turbo memory into a T60.

I don't think you can put turbo memory in T60 by any means.

[Tholek]

Wonder if it works if we swap turbo memory into a T60.

I don't think you can put turbo memory in T60 by any means.

It's a mini-PCI-E module, isn't it?

[cshbc]

If the WAN is a mini-pci-e, then that's what the turbo memory is. When you add the turbo memory to the build it warns you, that you cannot include a wan.

I did the search btw, and I must be a handicapped searcher b/c I could not find the above referenced thread by Anandtech.

I don't think you can put turbo memory in T60 by any means.

It's a mini-PCI-E module, isn't it?

[EOMtp]

It's a mini-PCI-E module, isn't it?

Yes, but the T60/p BIOS needs to support it, and it does not.

[RonS]

I've been look for an answer to the same question. It appears that Robson (Intel Turbo Memory) is a good idea that doesn't pan out in real life. Kind of like the promise of HyperThreading. After a lot of marketing hype, both of those technologies can easily end up slowing your system down rather than speed it up.

Windows already does a good job caching your hard drive data in system RAM, so the performance boost while you're up and running *should* be negligible (assuming that the system has a decent amount of RAM). The theoretical benefit comes during boot-up, where the startup data that is normally read from the hard disk is already pre-loaded into the Turbo Memory's non-volatile memory. But, from reading the reviews (primarily the one on Anandtech) it sounds like managing that procedure and/or non-volatile RAM performance is not up to par.

One possible benefit that I haven't seen discussed is power savings. If data can be read from and written to the Turbo Memory without having to spin up the hard disk, there should be some power savings. Again, the real-world benefit may also be negligable.

I ordered two T61 machines last night (at absolutely unbelievable prices) and I configured both systems without the Turbo memory. I figure I can always add it later if the technology gets ironed out.

[cshbc]

conversely then, will the t61 bios support mini pci-e WAN even if it was not a build option?

It's a mini-PCI-E module, isn't it?

Yes, but the T60/p BIOS needs to support it, and it does not.

[Tholek]

conversely then, will the t61 bios support mini pci-e WAN even if it was not a build option?

Yes, but the T60/p BIOS needs to support it, and it does not.

I'm thinking it probably could, since it's almost like a Wi-fi card, but then there's the 1802 error issue. It may apply here too.

[cshbc]

1802 or 1820 error?

conversely then, will the t61 bios support mini pci-e WAN even if it was not a build option?

I'm thinking it probably could, since it's almost like a Wi-fi card, but then there's the 1802 error issue. It may apply here too.

[Tholek]

1802 Unauthorized network card is plugged in. Power off and remove the Mini PCI network card.

That's what given when a compatible, yet "unauthorized" mini-pci card is inserted into many recent TPs.

Not saying that'll happen here, but I wouldn't be surprised.

[tomh009]

That would require a different BIOS build in the versions with and without the "turbo" memory, which is highly unlikely.

[EOMtp]

... will the t61 bios support mini pci-e WAN even if it was not a build option?

Yes.

[XIII]

The BIOS is more like direct-hardware access software implementation. More importantly, the current chipset in T60 does not support TM. So even if you force your T60 to accept it, it will be useless.

[spotter]

my guess is the bios only matters for bootup purposes. Once you're in Vista (or any modern OS) the bios isn't used. So if one would connect a turbo memory card in a T60, vista would probably see it, and be able to use it like it uses a usb key, but it won't make much of a difference.

but I could be talking out of my [censored].

[tomh009]

Assuming the T60's "Napa" chipset knows how to talk to such cards. The cache memory was advertised as one of the features of the new "Santa Rosa" chipset.

[XIII]

The current chipset used in T60 is dated back all the way to the released day of the Centrino Duo. Intel and Lenovo have not updated their chipset since then. Intel has not released Robson at that time yet. Asking for Robson support in the Napa platform would be impossible.

[spotter]

I'm unsure why you think napa has to be able to talk to the card itself for it to work in Vista. Napa knows how to talk pci express (it supports the WWAN card via it), what else would it have to know, what else does it have to do?

Again, Santa Rosa's feature is Intel's Robinson technology, not readyboost. Readyboost is just about while windows is already running, Robinson is about speeding up boot as well (before vista is loaded).

Basically, I don't get how robinson/readyboost could ever help significantly. No flash ram disk has the throughput of today's fast hard disks. The only advantage they have is basically minimal seek time. Especially in a boot scenario where its going to work the same way each time, an intelligent ordering of blocks on disk (so hard disk never has to seek) would seem to be a more optimal situation.

Basically, if I were doing this, I would force the disk to be partition to provide a "boot" area, not under control of the user as well as the normal disk. The OS would monitor itself on each boot to see the order that it seeks, it would then mirror the files to the "boot" partition and place their blocks in a fragmented way so that the OS doesn't have to seek (provided there's a decent amount of ram, and on most modern systems, that's a decent assumption, if there's a lack of ram, you're up a creek w/o a paddle already).

Each time the OS boots, it will monitor its seeks to optimize the boot area layout (for instance, you upgrade programs, drivers that are used during boot, that can cause them to read in different parts, but ordering will only be known after boot).

Until flash ram is has faster throughput than hard disks, I don't see how robinson technology could beat this. Similarly, if one has a decent amount of Ram in one system's I don't see how readyboost can help, and if one doesn't, upgrade instead of using a kludge

[brentpresley]

I'm unsure why you think napa has to be able to talk to the card itself for it to work in Vista. Napa knows how to talk pci express (it supports the WWAN card via it), what else would it have to know, what else does it have to do?

Again, Santa Rosa's feature is Intel's Robinson technology, not readyboost. Readyboost is just about while windows is already running, Robinson is about speeding up boot as well (before vista is loaded).

Basically, I don't get how robinson/readyboost could ever help significantly. No flash ram disk has the throughput of today's fast hard disks. The only advantage they have is basically minimal seek time. Especially in a boot scenario where its going to work the same way each time, an intelligent ordering of blocks on disk (so hard disk never has to seek) would seem to be a more optimal situation.

Basically, if I were doing this, I would force the disk to be partition to provide a "boot" area, not under control of the user as well as the normal disk. The OS would monitor itself on each boot to see the order that it seeks, it would then mirror the files to the "boot" partition and place their blocks in a fragmented way so that the OS doesn't have to seek (provided there's a decent amount of ram, and on most modern systems, that's a decent assumption, if there's a lack of ram, you're up a creek w/o a paddle already).

Each time the OS boots, it will monitor its seeks to optimize the boot area layout (for instance, you upgrade programs, drivers that are used during boot, that can cause them to read in different parts, but ordering will only be known after boot).

Until flash ram is has faster throughput than hard disks, I don't see how robinson technology could beat this. Similarly, if one has a decent amount of Ram in one system's I don't see how readyboost can help, and if one doesn't, upgrade instead of using a kludge

Current HIGH END NAND is approaching the burst transfer rate of high-end hard drives. Within 12-18 months, we will see NAND out in mass that puts HD transfer rates to shame. That is when Turbo Cache and SSD hard drives will REALLY shine.


Your analysis is a bit flawed because it only looks at ONE aspect of NAND products, the transfer rate. Right now, granted, that is not up to snuff with 7200RPM laptop hard drives. But what you are forgetting is the other VERY significant aspect to NAND-based products. ACCESS TIME.

Currently, the access time on NAND is almost two ORDERS OF MAGNITUDE faster than that on a hard drive (4-5ms vs 0.1ms). For small file block transfers, no hard drive can ever match the performance of this.


So think of TurboCache and Readyboost as a poor-man's L3 cache that is EXTREMELY efficient for small files. And inefficient for large files (for the moment).

[spotter]

actually I didn't

Basically, I don't get how robinson/readyboost could ever help significantly. No flash ram disk has the throughput of today's fast hard disks. The only advantage they have is basically minimal seek time

now lets go through your points. "High End NAND"

1) what costs are we talking about?
2) what does something that will be out in a year or two mean today? My analysis is about today

"access time" I did mention that, and I mentioned how one can avoid seek time with a hard disk, i.e. by not seeking, by placing the blocks on disk in the exact order they'll be read.

I don't see how readyboost is any better than the OS's own page cache.

[XIII]

If you read Anandtech review of Santa Rosa chipset, you will see Intel hinted that we should see more benefits of Robson.
IMO, Robson is meant to substitute for Hybrid Harddrive. Think about it, Hybrid hard drive is basically harddrive with large cache to reduce drive access. What if, Robson can simulate that large cache, we can experience better battery life and Hybrid harddrive perfomance on a normal drive.

[spotter]

Given today's technology, a hybrid hard drive only has minimal benefit during boot. Once the machine is booted, and the operating system's page cache takes over, main memory will be faster for caching than flash will ever be.

Then its just a function of size of main memory vs. size of working set.

[barrywohl]

Thanks for the discussion. It seems like Robson is a pretty good idea for some situations such as too little ram and too little money to buy more but, to date, no one has really demonstrated a benefit even in that situation.

It seems like speeding up boot up is a major advantage, but operating systems like VISTA don't need as many reboots as older systems such as Windows 95/98. Most of my reboots on my R50p with XP SP2 come from Outlook 2002 doing something quirky. Perhaps killing the outlook.exe process and relaunching would be enough.

I'm looking for more side by side real life experiences with Intel Turbo Memory in systems with 3 GB of ram. The two reviews I've read on line failed to show faster performance with Intel Turbo Memory.

Thanks to everyone who has contributed to this thread so far.

[wackydan]

Just remember that turbo Memory/Robson will only work under Vista.....

MS and Intel are not supporting it under XP. You'll Have an "unkown pci memory controller" as a device bang under XP.

[barrywohl]

As of right now (Sunday evening 3 June 2007) you cannot choose Intel Turbo Memory on the www.lenovo.com site for either the R61 or the T61.

[coolhandluke]

You have to choose at least 1GB RAM for Intel Turbo Memory.