Problems? See this thread at archive.org.
Do you have a link where I can get the Samsung 64gb SLC for $150?
-
This proves that the Anandtech article's author doesn't even bother sending Samsung an email asking whether they will provide a firmware update for Samsung drive or not, but pull a statement out of nowhere, without any basis, that Samsung won't update the firmware for their SSD to support TRIM.
I don't know about you, but I think there's something fishy going on here
And Lenovo also provide a similar firmware update for their Samsung SLC drives as well. There's no reason why Dell and Lenovo won't provide another firmware update.
Also, there was absolutely no issue with losing data when I updated the firmware of my Samsung drives. -
well, there were some drives on ebay, but, can't find them anymore,all sold out ?
-
-
-
-
sgilmore62, I have tried different ram disk products over the years and can honestly say that they promise much more than they deliver.
However, when I found that superspeed now offers a Win 7 x64 compatible version, I had to waste a few hours and try this once again.
The system is an 8GB Sony Vaio, 500GB 5400 RPM Scorpio Blue HD, running Win 7 x64. I installed the v10.1 Ramdisk Plus 64bit version and allocated 4GB of ram as an NTFS HD (the defaults). I pointed my temp file variables to the ram disk and also PS's scratch disk. I also put a 512MB pagefile on the RAM drive but Win 7 would blue screen when I shut off or rebooted the computer (if the computer went to sleep or hibernated, it was fine) so I returned the pagefile to C:\ and then ran some tests.
Of course, the Crystal Diskmark scores were insane, but like I've mentioned before, benchmarks don't do it for me, so I loaded a 2048x1536 pixel image in PS CS4 and for the sake of simplicity simply did these transformations to it:
Reboot the computer, wait 15 minutes for Windows to 'settle down' (Win 7 doesn't need 15 min, but I wanted to be sure it was 'settled') and do the following tests - once with the scratch disk on the RAM drive, once with it on the Scorpio drive (but leaving the RAM drive in place and still pointing the temp files to it) and finally with the RAM drive removed and using the full 8GB of RAM.
This was my simple test scenario:
1) Start PS
2) Load 3MP image (18MB)
3) Flatten image
4) Re-Size to 1000% (1.76GB)
5) Re-Size to 50% (450MB)
6) Rotate 90 degrees CCW
7) Re-Size to 50% (112.5MB)
8) Rotate Arbitrary 15.6 degrees CCW)
9) Exit PS
Yes; no filters, no fancy layers, nothing that would detract from seeing the effect of the RAM disk vs. no RAM disk.
The results were not too surprising to me, but they are worth mentioning; with the RAM disk, the operations completed at least 200% faster (43secs vs. 89secs) than when the PS scratch disk was pointed to the Scorpio Blue drive (both, with an effective 4GB of available system RAM).
(If I had waited until the HD activity slowed down with each step (when scratch=Scorpio drive) then the 'score' would be more like 1800% faster for the RAM disk, but although the disk drive light was lit almost constantly, it eventually allowed a click for the next step to (eventually) start).
The pro's to using a RAM disk as the scratch disk was that in addition to the minimum 2 times faster response vs. using the Scorpio as the scratch disk location, the system was always very fluid and never seemed to choke on the work it was doing.
This is in contrast to when the Scorpio was used - the system felt very 'disconnected' from the user - almost as if a virus had infected the computer (the mouse and/or keyboard would respond at very irregular and jerky intervals and the system would be unresponsive until the most intensive operations had completed (steps 4, 5, 6, and 7).
However, the cons of using 1/2 the 8GB RAM for a RAM disk were also a 5 minute reboot time (actually closer to 7-8 minutes until the system was 'usable') and at least an incompatibility for the pagefile (that would blue screen on shut down, when the pagefile was on the RAM drive).
Another major con is that PS will not 'release' the scratch file unless you restart PS. This is a major pain as when the scratch disk was full, there goes your speed advantage. With the above test scenario and using less than 30MB for the temp files the available space on the RAM disk was only 199MB free after I completed step 8 above.
That means that effectively, that's all I could do with this file! This is why I flattened the image (step 3), and why I resized to 50% twice - the limited size of the scratch disk would not complete the next commands if I didn't.
I must also state that had I used an action in PS to do the above steps (I did them manually), there would be much more than a 2 times advantage to the RAM disk (the faster you executed the commands on the Scorpio, the slower it would get because the HD would still be thrashing from the previous command). One more point for the RAM disk (not related to PS's speed increase) is that the system was noticeably more 'fluid' in general use simply because the temp files were pointed to it (maybe only 1% or 2% faster, but yes to me it's noticeable).
Finally, I put the scratch disk and temp pointers back to the Scorpio, uninstalled superspeed Ramdisk and rebooted the computer and re-ran the test.
Hmmm... yes, this is even faster by a few seconds over the RAM disk setup. More importantly, the reboots were now 1 minute - no incompatibilities / blue screen of death on shutdown/reboot, no limitations in PS because of a 'small' scratch file size (I could complete step 4 without doing step 3 and I could also complete step 6 without doing step 5) and although the temp files were being created on the Scorpio and the computer was a little less 'fluid' feeling, overall the computing experience as a whole was worlds better than with the RAM disk and an effective 4GB of system RAM.
So, I never reinstalled the RAM disk and saved myself $100. I figure that unless you have a system with a minimum of 48GB or more RAM installed, a RAM disk is actually decreasing your productivity (with my usage anyways).
(Yes, I do mean at least 48GB of RAM - I regularly use 50-60GB on my Raptor that is dedicated solely as the PS scratch disk file).
Of course, if you could afford a system with 48GB or more, then you'd already be using RAIDed SSD's and your system would already be going at Super speed with no need for superspeed.
So sgilmore62, unless you're just 'playing' with PS, I would say forget about ramdisks and just enjoy the 8GB RAM when you get it.
(You owe me a few hours for doing these tests for you)!
-
JW,
Did you see this?
http://www.techspot.com/review/201-ocz-agility-60gb-slc-ssd/
I'm sure this is the drive you're interested in.
What I find funny is that the article doesn't compare any Intel drives to it? Hmmm...
Myself, I would lean towards the Intel (as everyone here knows, ) mostly because of the slightly higher capacity in this case but up to half the price (when you can find one in stock, of course).
If your specific requirements for the SSD require 24/7 operation and/or the highest read AND write speeds, then the SLC is your 'only' choice especially at that price.
The higher capacity is not just a 'need more storage' requirement - it can actually make a difference in how the SSD performs (the higher the capacity, the faster the drive is - generally speaking) so keep this in mind too when you make a decision. Leaving at least 20 to 30% free is a good idea for any hard drive -especially your main O/S drive. -
-
Quote:
Originally Posted by ronan_zj
I just got an email from Samsung COrp,
here it is:
zephir, hehe, sorry, but ronan_zj just got that email yesterday, Anand's article is dated August 30th.
Even then, what you state he said and what he's actually said (I've just re-read the article yesterday) do not agree. I would love to see a high profile cover up of Intel 'antic's', but this simply does not qualify as one.
Sorry.
-
JW,
For your use, I would seriously reconsider the Intel 160GB G2. Not only is the Intel rated for 100GB writes per day (no typo) for minimum of five years, it is actually optimized for workstation use as your description seems to indicate (vs. the enterprise target market of the SLC based drives which focus on IOPS specs along with long term reliability).
As to the G-Monster - don't be a sucker!
Remember that your system is limited by the Sata2 spec's and currently that is around 260MB/s effective throughput. Anything that promises more than that on a single Sata2 channel is lying.
Also, is your HP system Sata2 or Sata1? If it's Sata1, then you're further 'artificially' limited to around 130MB/s effective throughput.
If you haven't already, I would invest in maxing out my RAM to 8GB or more, and/or buying the 160GB Intel, use it till it dies, then upgrade to the latest and greatest (around 2014!).
Yes, SLC's are 'the very best', but for your usage scenario, you can be using several versions of 'only' current 'bests' that will probably surpass today's SLC's in one or two versions/years at the most.
As to SLC's are the 'safer' way to go, hmmm, I would say Intel is the safer way to go (whether it's their SLC's or their MLC drives) with all the information I've devoured on this SSD topic.
My advice? Buy the best, but also buy wisely.
-
-
-
-
I don't mention, nor imply any under-table dealing between Intel and the Anandtech writer. I'm simply saying that it's quite fishy that if the writer sends an email to Samsung, undoubtedly he would get an answer regarding whether Samsung will release a TRIM-supported firmware update or not, very similar to what ronan zj does. In fact, he may even include that email in as proof in the article. Instead, he states that you need to send the drive to Samsung to get it updated, without referencing any press release or piece of information directly from Samsung on the matter.
And yes, I did read the article
-
-
-
JW, I think you lost me.
Let me try a different explanation;
Yes, the G-Monster may be able to process internally at 500MB/s, but what good does that do when you can only access that data at 260MB/s max? Answer: it does no good. Unless you value transferring huge amounts of data all day (because it's sustained sequential write speed should be below or very close to the effective 260MB/s too).
If you RAID0 two SSD's (even if you use software RAID or the mobo's RAID) on two separate SATA connections, then you will definitely see a performance increase over one SSD.
Finally, even if you only have a Sata1 connection for your SSD, you will still see a huge increase in performance over the mechanical hard drive because the write speeds (and especially the 4K random read and write speeds) will still be slower than the maximum the Sata1 connection will allow.
However, if you do have a Sata1 connection and you opt for an SLC drive, you are basically throwing away almost 50% of it's performance in the one spec that it really matters in a workstation type scenario; it's 'sequential write' performance.
So, again, your call - but first do you have a Sata1 or Sata2 capable computer?
(If you have Sata2, then the '$$$pain' of owning an SLC drive is more bearable!).
Hope this clarified it a little for you? -
-
Oh! I would've jumped at that Samsung deal!
Maybe I'm too old and cautious, but I would be very weary of flashing a firmware upgrade for anything that was not made specifically for my hardware.
Flashing a Dell or Lenovo SSD firmware update onto an non-Dell or non-Lenovo computer seems like you like to play with fire?
There are simply too many things interacting that might go wrong if given half a chance that I wouldn't consider doing it myself (well, at least not on my 'work' systems). I'm glad it worked for you, but I don't think I would be so brazen.
Just curious as to how you would know if it's (the firmware) not brand specific? Thanks! -
-
JW, I use Tom's charts all the time, yes they're awesome for recommending parts to others (myself, I am usually into the new stuff that the charts are behind on!).
Now, the G-Monster is basically two SSD drives put into one chassis and uses a RAID0 controller to enhance it's speed. What is mostly enhanced is it's write speed as this is where most SSD's stumble. For a desktop system, it is more economical to simply use two SSD's to achieve the same thing (but save yourself the price of the G-Monster's RAID controller as you can use most mobo's built in RAID support).
For a Notebook system, we may not have the luxury of using two SSD's and even if we can, we can't RAID them (easily). As I mentioned before, whether the G-Monster is a viable option for you depends on your use/needs of the SSD and also the notebooks Sata mode (1 or 2).
If you are highly dependent on write speeds in your applications, then even with 'only' a Sata1 interface the G-Monster may still be the most desirable SSD. However, there are not many real world uses of an SSD's and a notebooks limited capacity along with the high (highest) write rates.
If I am understanding your 'theory' above, (and I admit I don't think I really do) then we must understand the flow of data from the processor to the final storage medium.
For any work to be done, it must be done by the processor. Data must always first pass through the processor for us to see or effect a result. Second, data is only able to be worked on as long as it is in physical RAM. Now, whether we save the data to a mechanical hard drive or to an SSD, it is simply stored there until we load it up in RAM to view/edit again.
Again, I emphasis that the SSD or the mechanical HD is simply 'dead' storage.
Knowing the above, can you now see that it doesn't matter how fast the G-Monster is internally? Although it has a RAID0 controller deciding how the data is stored, that's all it knows - to store data. It can't do 'work' on any data, nor can it change the data by itself (actually this would be called a data error if it did!).
So, as fast as it works internally, the only criteria that determines it's effective or 'actual' speed is the speed of the connection (Sata1 or Sata2) it can communicate with the cpu and/or RAM with. So in effect, the connection determines the maximum speed.
Once Sata3 is available, then the 'real' speed of a G-Monster may be known, but right now, it's only an intellectual curiosity unless you need and can really use the highest write speeds possible within a single SSD.
Even then, I still would choose an Intel SLC!
(Not quite sold on the 'no returns/no refunds' policy of the DVNation products).
Let us know how your G2 performs and if it surpasses your expectations. Hmmm... I really should buy some Intel stock, eh?
-
-
JW, Thanks!
Is that your notebook in that link?
I'm not laughing at you at all, but to continue your car analogy, even when you get the motor at 800+ HP, you still have to find a way to cool it AND connect all that power to the road.
To 'cool it' is easy to understand, to 'get all that power to the road' is to remove as many barriers/obstacles from the CPU as you can - and SSD's go a long way to achieve this goal.
When as many obstacles are removed from the cpu's way, then the only way to make the computer even faster is to get a bigger, badder meaner cpu (and i7 is certainly the way to do that!).
Take care. -
Lenovo has been known to push out firmware for all conventional HDDs or SSDs a few days after they're released. Based on my update schedule for normal HDDs, I took a chance and use the SSD update on a OEM Samsung SSD and it worked perfectly. Been using the drive for 3 months after the update, and it's still as good as day one
-
The new SSD Thread (Benchmarks, Brands, News and Advice)
Discussion in 'Hardware Components and Aftermarket Upgrades' started by Les, Jan 14, 2008.
