Re-visiting HDDs.

[VirtualLarry]

In my most recent build, a G3258 @ 4.2Ghz, 2x4GB of DDR3, on a Gigabyte H81 board, I used a 160GB WD Blue HDD 7200RPM drive.

It took noticeably longer to install, almost painfully so.

And it takes time to boot, so much that you would prefer not to boot very often, but it's not quite to the point that I would call it "excruciating".

And things like Windows Updates pin the Disk at 100% for extended periods of time, slowing down opening other programs.

Scans like Malwarebytes take 5min, rather than less than 1min.

But on the other hand, it cost me $12, and not $60.

Tradeoffs...

What really opened my eyes, though, is the Review that someone posted in my recent Biostar SSD thread. On one of the later pages in that review, it has a graph of file-copy speeds.

And the WD 1TB Blue HDD 7200RPM, is FASTER than some of the other known-crappy SSDs!

Which kind of opened my eyes.

HDDs are not optimal for multi-tasking. But modern HDDs aren't really that "slow" anymore, especially with the greater RAM cache amounts on them.

Sure, a GOOD SSD is faster than a GOOD HDD, but it's equally possible that a GOOD HDD can be faster than a mediocre / poor SSD, for certain things.

Oh, my point. That after using it for 30 minutes with Firefox 58a1 Nightly, I really had a hard time seeing that it was that much slower than an SSD.

Sure, if you're opening up a massive array of tabs when your browser opens, then an SSD will have a definite advantage.

But once the browser is loaded into RAM, and with a fast internet connection (I have gigabit), unless you run out of free RAM and start thrashing the pagefile (a real and present danger to using a HDD for browsing), then things aren't that much different, browsing individual pages, between a relatively-fast HDD and an SSD.

 

[cbn]

When I did my test with 2 x WD5000AZLX (1 TB platter with 32MB cache) in RAID-0 as a boot volume I was quite surprised at how fast they were.

Haven't tried a single modern HDD for OS yet though. I would imagine it would do well though.

P.S. This hard drive, in contrast, became painfully slow (over time) with anything other than web browsing after I began using the PC it was in as a free software test mule.

 

[XavierMace]

Larry, just once, I'd love to see one of your threads like this have some actual numbers rather than just perceptions. LOL.

 

[VirtualLarry]

Well, I'd probably have to record some videos to capture the essence of what I've been seeing. Which might happen, once I dig out of my debts and get a decent video camera so I can start doing YouTube vids.

 

[dave_the_nerd]

The “some things” where a top-rung hdd is actually faster than the crappiest of crappy SSDs are limited to sequential transfers in synthetic benchmarks. Basically no real-world workload is like that.

In other words: No. Nope, nop, nu-uh, no way. Don’t, nope, stop, no halt. Never no nein nada, no no, stop no. No, and no. Don’t even think it, don’t say it, not one fragment of a single thought that has anything to do with his. No.

Might as well boot your system from a Linux LiveCD with a 4x SCSI CD-ROM. Just because it’s cheap & technically works doesn’t make something a good idea.

 

[dave_the_nerd]

Well, I'd probably have to record some videos to capture the essence of what I've been seeing. Which might happen, once I dig out of my debts and get a decent video camera so I can start doing YouTube vids.

Or you could just time some common tasks with a stopwatch.

The early Hyperthreading benchmarks that Anandtech did back in the late Pentium 4 era had some multitasking benchmarks you could probably duplicate easily.

My personal favorite is to run something like Geekbench, and then run it while a virus scan is running, and observe the difference.

You can also whip up your own benchmark utility fairly simply in Python.

 

[BonzaiDuck]

1) I have a place in my systems for at least one large-size HDD. [To me, "Large" can mean >= 2TB].

2) I don't "do" RAID0 anymore.

3) Larry and everybody else can guess what I'd add for "4," "5," etc. etc. and I won't pursue it.

. . . Except to say that my more expensive experiments on one system and a similar cheaper strategy on a second system give me some really good benchmarks . . . But it still costs me extra for more RAM, and extra for an additional solid-state storage device. And . . . some software, of course . . .

 

[Insert_Nickname]

What really opened my eyes, though, is the Review that someone posted in my recent Biostar SSD thread. On one of the later pages in that review, it has a graph of file-copy speeds.

And the WD 1TB Blue HDD 7200RPM, is FASTER than some of the other known-crappy SSDs!

Which kind of opened my eyes.

You are being blinded by sequential R/W speeds. Which on newer HDDs are decent'ish. Older HDDs are another story entirely.

The strong point of SSDs are, and have always been, random I/O. An area where even 15000RPM HDDs can't keep up with even the cheapest SSDs. On your typical 7200RPM HDD a random request will take on average 4.17ms. If you're unlucky the R/W head has just moved past, it can take up to 9ms. Full stroke search can take 20ms. Best case track to track search is typically 0.5 to 1ms. Milliseconds are an eternity in a PC. Just think what that kind of access time does to the swap file.

In comparison SSDs typically have sub 1ms latency for all R/W operations. If they don't, that model should be avoided.

In other words, you can pull perhaps 100-150 IOPS from a 7200RPM drive. Even my ancient 40GB X25-V could manage 1000+, and was a lot snappier then even the fastest HDDs then.

Oh, and large caches don't help if you're dealing with an SMR drive. They need that cache, and even a PMR cache, just to give acceptable performance in the first place.

Sure, a GOOD SSD is faster than a GOOD HDD, but it's equally possible that a GOOD HDD can be faster than a mediocre / poor SSD, for certain things.

Oh, my point. That after using it for 30 minutes with Firefox 58a1 Nightly, I really had a hard time seeing that it was that much slower than an SSD.

Sure, if you're opening up a massive array of tabs when your browser opens, then an SSD will have a definite advantage.

But once the browser is loaded into RAM, and with a fast internet connection (I have gigabit), unless you run out of free RAM and start thrashing the pagefile (a real and present danger to using a HDD for browsing), then things aren't that much different, browsing individual pages, between a relatively-fast HDD and an SSD.

When everything is already loaded into RAM, and you have enough of that RAM, sure, you won't see much performance difference. But start anything that isn't loaded, and you will.

Back in the day you could totally freeze up a PC by filling the available memory, and then starting something that required the swap file. Double whammy disk thrashing...

 

[cbn]

Well, I'd probably have to record some videos to capture the essence of what I've been seeing. Which might happen, once I dig out of my debts and get a decent video camera so I can start doing YouTube vids.

Best way (I think) would be to use some screen capture software---> http://www.techradar.com/news/the-best-free-screen-recorder

Fancy showing other people how to play big name games on YouTube? Would you like to make demos to show off your apps, or to teach others how to use specific software? Then you need a screen recorder, and there are some excellent options available completely free.

The simplest screen recorders simply capture what's on your screen and save it in AVI format, but the more advanced free tools also offer editing, picture-in-picture, live streaming, and gameplay optimization.

P.S. I was thinking of doing the same thing with my hard drives in RAID-0 back when I wrote this post.

 

[cbn]

Something else to think about would be SSHDs like the one you posted here.

I only wish they had more than 8GB NAND.

I suspect part of the reason Seagate dropped the 32GB versions was due to poor synthetic benchmarking numbers, see example of SSHD with 32GB NAND in the review below:

https://www.back2gaming.com/reviews...views/seagate-laptop-sshd-st1000lx001-review/

The new Seagate Laptop SSHD again drops us in a tough spot wherein benchmarking the drive presents an absolute challenge as its technology is not even remotely close to fully benchmark-compatible.

See your benchmark of a 32GB Kingfast F6 SSD as reference point:

https://forums.anandtech.com/thread...-128gb-ssd-drops.2467389/page-2#post-38272623

The reason, of course, for the lower than expected sequential read is that CrystalDiskMark is not likely to be found in the NAND cache of the SSHD.

 

[cbn]

Which might happen, once I dig out of my debts and get a decent video camera so I can start doing YouTube vids.

4K or 1080p?

Interestingly enough, 4K is getting fairly affordable now with the Panasonic G7 ---> http://learnaboutfilm.com/panasonic-g7-best-budget-camera-filmmaking/?utm_expid=.zQC7ujJNQi-5KYVxHke6Bg.1&utm_referrer=http://learnaboutfilm.com/making-a-film/equipment-for-low-budget-filmmaking/choosing-a-system-camera/ (NOTE: The camera does have a microphone jack and can be used with an external microphone like the Rode VideoMic).

In fact, here is a pretty nice package deal (G7 Camera plus 7 accessories, including a 64GB U3 SD card necessary for recording 4K30 video ) for $600.

EDIT: Interesting video called "Should you make 4K videos for Youtube yet? (Panasonic G7 used for the video)

EDIT2: Also consider the iPhone* (Some Lifeline providers like Safelink will let a person use one of these on their plan).

*Will also do some research on the iPhone 5s (this phone, unlike the iPhone 5 and 5C is still supported by Apple) and iPhone 6.

But since we are talking about hard drives in this thread, It's too bad Storage for 4K editing needs to be 4x more capacity and 4x faster than 1080p. (For hard drives the recommendation I have seen for 4K editing is at least three 3.5" in RAID-0 (or a single SATA SSD)....whereas 1080p can be handled with a single modern 3.5" hard drive.). For this and other hardware reasons (Monitor and Video card) I will probably focus on 1080p (and lower) for the foreseeable future.

 

[dullard]

Sure, a GOOD SSD is faster than a GOOD HDD, but it's equally possible that a GOOD HDD can be faster than a mediocre / poor SSD, for certain things... things aren't that much different, browsing individual pages, between a relatively-fast HDD and an SSD.

And you have stumbled upon the proper use of the first generation Optane drive. A good HDD can be fast in many applications and usually is quite large. Slap on an Optane for ~$40 (which should last you through many computer builds, so the cost is minimal once you spread it out) and then the line is blurred even further. If you need a SSD for some applications, use that HDD+Optane for fast Windows boot and the SSD as your secondary drive.

 

[BonzaiDuck]

And you have stumbled upon the proper use of the first generation Optane drive. A good HDD can be fast in many applications and usually is quite large. Slap on an Optane for ~$40 (which should last you through many computer builds, so the cost is minimal once you spread it out) and then the line is blurred even further. If you need a SSD for some applications, use that HDD+Optane for fast Windows boot and the SSD as your secondary drive.

Again, I could stand corrected or re-directed -- but whether or not you choose Optane with the HDD is not critical to the outcome, except that the Optane drive is persistent between system reboots compared to scenarios like the one that generated this benchmark:

2.5" 2TB Seagate Barracuda cached to 4GB of system RAM with PrimoCache:

Even so, the cache can be saved at shutdown or restart as a configuration option using only RAM. Otherwise, you could cache the spinner to an NVME caching volume.

How much Optane nets you in a similar configuration as an advantage over pre-NVME technologies depends on performance characteristics of 3D-XPoint primarily, and you wouldn't feel deprived without the ability to use Optane.

Note also my benchies show this working without deferred writes. With the greater risk that deferred writes brings, the sequential and random write rates appear astounding, as much as benchmarks are as much "appearances."

 

[dullard]

Again, I could stand corrected or re-directed -- but whether or not you choose Optane with the HDD is not critical to the outcome, except that the Optane drive is persistent between system reboots compared to scenarios like the one that generated this benchmark:

16 GB of DDR4 is ~$80 vs 16 GB of Optane for just over $40. Plus, the RAM needs to be replaced with most motherboard upgrades, while one Optane can just be moved from computer to computer over generations.

RAM drives are great and terrible. I used to use them a lot when it would take eons to save files. It would save tons of frustration. I would save frequently to a RAM drive and save every few hours to the real drive. That frustration of slow drives is gone, that is, until power is lost unexpectedly.

 

[BonzaiDuck]

16 GB of DDR4 is ~$80 vs 16 GB of Optane for just over $40. Plus, the RAM needs to be replaced with most motherboard upgrades, while one Optane can just be moved from computer to computer over generations.

RAM drives are great and terrible. I used to use them a lot when it would take eons to save files. It would save tons of frustration. I would save frequently to a RAM drive and save every few hours to the real drive. That frustration of slow drives is gone, that is, until power is lost unexpectedly.

"RAM drives are great and terrible . . . " Often, I see that their is a misconception or confusion about what is discussed. Romex Software produces both "RAM DISK" and Caching software. Their "Primo" product is a Swiss-Army knife of RAM and SSD-caching options, so offering the user the equivalent of both Samsung "RAPID" and Intel "ISRT" without any proprietary limitations. It also allows simultaneous caching of AHCI and RAID-mode disk configurations, caching across different controllers, etc.

I really don't want to be the forum "Hot-Dawg" promoter of this or that software, but I was seriously investigating the existence of such software back in 2014, saw how some of the players would get bought out by disk manufacturers or the software was not kept current with new OS versions. The only two options I found were PrimoCache and Superspeed's SuperCache . I see that SS has updated its web-page presentation, and I may want to look at it again, but it offered much less in options than Primo. When I looked more closely at it in the past, it didn't offer persistent SSD caching.

There are skeptics. Most of the folks at the Romex forums are happier than pigs in s*** about it, and you always see some neophyte or noob coming in there asking questions and emerging with conviction.

Anyway, I'm using the Z170 chipset. I don't think Optane "fits" on the Z170, or it isn't offered as an option -- correct me if wrong. But if you don't have the option, $80 versus $40 in a scenario of possible motherboard upgrade (for -- what? -- $140?) is chump-change. Of course, the software for lifetime license is about $30.

On the RAM issue for DDR4. I agonized for months over upgrade options of 16GB to 32GB, and the preference for replacing 2x8 with 2x16GB. Rumor had it that you couldn't just add a 2x kit of the same model and speed and fill all slots. Even G.SKILL was telling me they couldn't guarantee that a 2x (2x8GB) configuration would work at full spec, if it wasn't a 4x8GB matched kit. But I went ahead and bought a second 2x8 kit, and it works great at spec speed and timings (3200, 14-14-14, etc.). You just can't set CMD = 1 with DDR4 for four modules and expect success. I'll only close by saying someone else may not have such good luck.

Also, I may have indicated because of the focus our topic had here, you can use Primo to cache either SATA HDD or SSD to an NVME SSD. How much better it gets with Optane would still likely be somewhat marginal, even with "3D Xpoint."

 

[cbn]

Windows (on my computer) appears to use RAM fairly aggressively for Read cache. (I have 64GB of system RAM and task manager (under the performance tab) usually shows a very large amount of this 64GB being used for cache if I leave my system on long enough).

So I have been thinking about Romex Primocache using the part of the system RAM as a write cache? This particularly for someone with a hard drive as the boot drive.

P.S. Would be interesting to see how Romex Primocache works on a laptop (as a write cache to DRAM) with either a 2.5" Hard drive, a Firecuda 2.5" SSHD or a 2.5" hard drive (coupled to 16GB Optane) as boot drive. This especially as the battery of a laptop essentially is a built in UPS.

 

[cbn]

And you have stumbled upon the proper use of the first generation Optane drive. A good HDD can be fast in many applications and usually is quite large. Slap on an Optane for ~$40 (which should last you through many computer builds, so the cost is minimal once you spread it out) and then the line is blurred even further. If you need a SSD for some applications, use that HDD+Optane for fast Windows boot and the SSD as your secondary drive.

It certainly does benchmark well.

http://www.thessdreview.com/our-rev...memory-module-review-32gb-every-pc-user-know/

1TB 3.5" WD Black Drive Hard drive by itself:

1TB 3.5" WD Black Drive Hard drive + 32GB Optane:

 

[BonzaiDuck]

Windows (on my computer) appears to use RAM fairly aggressively for Read cache. (I have 64GB of system RAM and task manager (under the performance tab) usually shows a very large amount of this 64GB being used for cache if I leave my system on long enough).

So I have been thinking about Romex Primocache using the part of the system RAM as a write cache? This particularly for someone with a hard drive as the boot drive.

P.S. Would be interesting to see how Romex Primocache works on a laptop (as a write cache to DRAM) with either a 2.5" Hard drive, a Firecuda 2.5" SSHD or a 2.5" hard drive (coupled to 16GB Optane) as boot drive. This especially as the battery of a laptop essentially is a built in UPS.

In 2014, the trial version ran for 90 days. It now gives you 60 days. There are options for different cache usage, "Read," "read and write", etc. etc. I use it on a laptop currently, but that's the point at which I started with it after upgrading a 2008 Gateway to 8GB SO-DIMM. So I'd suggest you download and install it to see what it does for you.

Meanwhile, keep an eye out for version 3.0. In fact, you might be able to get that one -- don't know what restrictions there are on Beta 3.0, whether available for general public or licenses 2.7 users. That's all I can tell you.

Also, per your post ending in " . . . sure does benchmark well . . . " for Optane + a WD HDD -- that's the sort of thing you can expect. It only confirms for me how Optane is a non-essential hardware option for me, given the other hardware (SATA and NVME) and a $30 lifetime-license software package. Or -- at least that's my perception at this time.

I'm just curious how the orchestration of Optane hardware and Intel software will shake out, or whether it's limited to that sort of combination.

It just seems a bit ironic with my recollections of "caching-skeptics" going back to 2011 with ISRT, and the mixed responses to ISRT, RAPID, and this "PrimoCache subculture." I mean . . . whatever hardware possibilities one has at the moment, one would try to kloodge together a combination that serves that moment. Suddenly, a niche piece of hardware appears from Intel on high, and there is a flurry of interest.

 

[cbn]

From the Primo cache overview:

• If you have 4GB or more memory, you can use a portion of the memory as cache to accelerate local disks, including mechanical disks, SSDs and flash drives, and iSCSI disks.
• If you have memory not seen by Windows, usually in 32-bit Windows, you can use this hidden memory as cache to accelerate local disks and iSCSI disks.
• If you have a SSD and a mechanical disk on your system, you are able to partition a portion or all of this SSD as cache to accelerate the mechanical disk.
• If you have a spare USB 3.0 flash drive, you can use it as cache to accelerate your mechanical disks.

Interesting that a usb 3.0 flash drive can be used to accelerate disks.

 

[cbn]

Small RAMDisk (List of free versions here) for browser cache? I wonder how much this could help speed up a single hard drive used for OS?

P.S. Check out the inteface on ASRock XFast RAM:

Unlike the Asus RAMDisk utility, ASRock XFast RAM does not work on any computer. Therefore, you will need an ASRock manufactured motherboard to use it. There are switches to automatically send Chrome, Firefox, Opera and IE browser cache files to the drive, as well as setting it up for system temp files, the system page file, WinRAR temp files and even Ready Boost cache space. The RAM Disk space is not artificially limited and can be as big as the amount of free RAM you have.

 

[cbn]

Here are some instructions (web page and video) on how to use the AMD Radeon RAMDisk with browser caches:

http://www.radeonramdisk.com/files/QuickUseGuide_MovingBrowserCache_V1.pdf

Some info on the AMD Radeon RAMdisk taken from the article linked in post #20.

As we mentioned earlier, Dataram RAMDisk and AMD Radeon RAMDisk are in essence the same program. The only major differences are Dataram has a smaller maximum RAM drive size of 1GB, you can select a FAT or unformatted filesystem and there is an option to choose a boot sector type. The AMD tool has additional options to choose what type of disk to load on startup and which drive letter to assign the RAM drive to. Both periodically popup an ad window.

 

[BonzaiDuck]

From the Primo cache overview:



Interesting that a usb 3.0 flash drive can be used to accelerate disks.

Here I'm responding to your remark above and the subsequent posts. I don't even know if I'd bother with a USB 3.0 flash drive, but maybe it would be worth trying just to "see." Perhaps it's fast enough to substitute for an SATA SSD cache. That actually would simplify some configurations, or keep them simple and free up an SATA port.

Those RAM disk options are interesting. I guess I'm reacting to it given my current position in some kind of configuration "checkerboard." With what I have, I wouldn't want to introduce additional complications, even as we might suspect they'd all work out properly.

Based on an exchange recently on the Romex forum, I'm changing my allocation of RAM to try a two-tiered approach for caching an SATA SSD to NVME and RAM. I decreased my cache for the NVME boot drive to 12GB, and gave the SATA SSD with NVME cache the 4 GB of RAM removed from the boot-system disk cache. It's supposed to work properly, so I'm going to try it again.

The benchie sequential read results after the change for the C: boot drive and the E: drive are approximately equal now at 13,000+ MB/s.

But ask me if I notice a difference over native Sammy 960 performance, I can only say "in some ways at some times." REally, the only thing I can say is "my system is not slow." I've never seen anything faster. I don't have any basis for comparison except with the other systems I've built, benchmarks posted by others, etc. Works for me, though!

 

[cbn]

Looking at Newegg I was interested what New desktop had the largest HDD for primary storage.

Well it turns out that 3TB is actually the largest currently on Newegg, this with 32GB Optane from CyberPowerPC :

https://www.newegg.com/Product/Product.aspx?Item=N82E16883230241&ignorebbr=1

(I couldn't find any desktops with 3TB HDD as primary storage without Optane).

Interesting, as that is only 1TB more than what we have seen in the budget category (which, of course, doesn't have Optane)---> https://forums.anandtech.com/threads/the-pre-built-atx-tower-hot-deals-thread.2492098/ ($400, Core i5 6400, 8TB, 2TB HDD, 300W ATX PSU, Windows 10 Home) EDIT: The current model with Core i5 7400 only has 1TB.

P.S. I did find a Gaming system from iBUYPOWER with 2TB HDD as primary storage using 16GB Optane Here. (None of the systems on Newegg with 1TB HDD had Optane)

 

[cbn]

Here are some instructions (web page and video) on how to use the AMD Radeon RAMDisk with browser caches:

http://www.radeonramdisk.com/files/QuickUseGuide_MovingBrowserCache_V1.pdf

Some info on the AMD Radeon RAMdisk taken from the article linked in post #20.

Yesterday I set up this RAM Disk and followed the videos instruction. I am already at the maxium 1GB capacity Firefox 57 Developer edition allows for cache. Do any other browsers allow for a greater amount of cache?

P.S. I am using this with a PNY CS2211 SSD (Phison S10 and 15nm MLC) so my browsing was pretty fast already. Eventually I will try this out with a hard drive (I have 2TB 2.5" Firecuda on the way but I will also use with an older 3.5" hard drive as well).

 

[VirtualLarry]

Interested to see how web browsing is affected performance-wise by using a RAMdisk for the temp/cache directory.

Not sure 1GB is enough though, for my usage.