Jump to content


 


Register a free account to unlock additional features at BleepingComputer.com
Welcome to BleepingComputer, a free community where people like yourself come together to discuss and learn how to use their computers. Using the site is easy and fun. As a guest, you can browse and view the various discussions in the forums, but can not create a new topic or reply to an existing one unless you are logged in. Other benefits of registering an account are subscribing to topics and forums, creating a blog, and having no ads shown anywhere on the site.


Click here to Register a free account now! or read our Welcome Guide to learn how to use this site.

Photo

Performance HDD is killing my Benchmarks....what can I do?


  • Please log in to reply
30 replies to this topic

#1 cat1092

cat1092

    Bleeping Cat


  • BC Advisor
  • 6,987 posts
  • OFFLINE
  •  
  • Gender:Male
  • Location:North Carolina, USA
  • Local time:12:00 AM

Posted 07 March 2018 - 12:23 AM

Topic Header fairly sums up question. I benchmark my machines on the popular UserBenchmark site, link at bottom of post, (I believe) my HDD, a 2TB WD Gold, a Performance/Prosumer drive performs like crap, the only weak link shown. 

 

I've added the HIPM-DIPM utility to the registry to prevent head parking, set to Active & 0 (this has been recommended since at least the early Windows 7 days, see below link).. Power Plan set to Bitsum High Performance, with Performance Mode enabled (ProBalance disabled for these test runs). 

 

https://www.tenforums.com/tutorials/72971-add-ahci-link-power-management-power-options-windows-10-a.html

 

Speccy specs of PC below, if more is needed, feel free to ask. 

 

http://speccy.piriform.com/results/ePNX3cDIz2gmvJsoQF1HmUX

 

The issue seems to be HDD related, although overall performance is Outstanding, am in the 10th percentile (leaving 90% somewhere). Have tried swapping ports, installing latest SATA drivers, disabling Intel Rapid Storage Technology & under the Device Manager, allowing for write caching, both results in disqualification if used. The drive is SATA-3 with a 128MB buffer, spinning at 7200 rpm & the replacement for the WD RE4 lineup. Am open to trying other drives as a temporary measure, yet this one is needed for the machine, cannot shrink & have the space needed. The WD Gold is considered a premier HDD, am on the low end of the spectrum at 2TB, larger one has a 256MB cache. Which I assume is supposed to make things faster, not the other way around. 

 

If anyone needs more data from me, other tests ran, whatever needed to diagnose situation, feel free to ask, am in 100% cooperative mode to anything to get this drive in the top 10%, rather than the last. Am near UFO rankings in the other two categories, I feel for the desktop to break the mark, the drive needs to be fast. Not having a HDD counts against those testing, so not the answer. :)

 

http://www.userbenchmark.com/UserRun/7679301

 

This low HDD ranking baffles me & I'll not rest until fixed. :P

 

Note to Moderators/Other Senior Staff: If this Topic isn't where should be, please feel free to move as necessary. :)

 

Cat


Edited by cat1092, 07 March 2018 - 12:34 AM.

Performing full disc images weekly and keeping important data off of the 'C' drive as generated can be the best defence against Malware/Ransomware attacks, as well as a wide range of other issues. 


BC AdBot (Login to Remove)

 


#2 MDD1963

MDD1963

  • Members
  • 688 posts
  • OFFLINE
  •  
  • Local time:01:00 PM

Posted 07 March 2018 - 01:55 AM

I would expect anyone using a conventional hard drive for an OS drive to be in the bottom 10%, unless the results are categorized to have spinning hard drive users benching only against each other...

 

In a pure 'storage performance' benchmark, hard to imagine the NVME M.2 (960 Pro/EVO) solutions being the top 10%, conventional recent generation (850 Pro, etc) proving the middle 80%, and, the 'spinning rust' conventional drives bringing up the rear...

 

Even if the results were categorized by types of drives, there are bound to be a handful of folks running a pair of whatever is fastest in RAID 0.....


Asus Z270A Prime/7700K/32 GB DDR4-3200/GTX1060


#3 Joe C

Joe C

  • Members
  • 775 posts
  • OFFLINE
  •  
  • Local time:12:00 AM

Posted 07 March 2018 - 06:49 AM

The max data rate for that drive is 200mb.

https://www.wdc.com/content/dam/wdc/website/downloadable_assets/eng/spec_data_sheet/2879-800074.pdf

 

You can check to see if the drive is aligned (because this is not being used as an o.s. drive)

https://www.seagate.com/tech-insights/advanced-format-4k-sector-hard-drives-master-ti/


Edited by Joe C, 07 March 2018 - 07:04 AM.


#4 jonuk76

jonuk76

  • Members
  • 2,140 posts
  • ONLINE
  •  
  • Gender:Male
  • Location:Wales, UK
  • Local time:05:00 AM

Posted 07 March 2018 - 08:59 AM

I think that particular disk benchmark is probably not the best anyway.  Look at the variance on the 4K (small file random access) test, against the same model on other computers.  And even the transfer rate test seems questionable as well, with most of the top WD Gold 2 Tb having a measured write speed of 240-250 MB/s, but much lower read speeds, which I find confusing.

 

I do wonder, if you have a drive mostly full, with lots of small files on it, is that going to count against it in such tests?  Particularly the random access tests.  Many of those benchmarking appear to be doing so with nearly empty drives.


7sbvuf-6.png


#5 DavisMcCarn

DavisMcCarn

  • Members
  • 732 posts
  • OFFLINE
  •  
  • Gender:Male
  • Local time:12:00 AM

Posted 07 March 2018 - 03:44 PM

Short of replacing the 2TB drive with an SSD, the only thoughts I have would be to uninstall Superantispyware which may be intercepting all of your disk traffic, disabling Skype and Spotify, then running the benchmarks again.


Computer dinosaur, servicing PC's since 1976

#6 Platypus

Platypus

  • Moderator
  • 13,700 posts
  • OFFLINE
  •  
  • Gender:Male
  • Location:Australia
  • Local time:02:00 PM

Posted 07 March 2018 - 05:09 PM

I do wonder, if you have a drive mostly full, with lots of small files on it, is that going to count against it in such tests?  Particularly the random access tests.  Many of those benchmarking appear to be doing so with nearly empty drives.


Yes, the notes for that test state the best results will be from an empty partition at the start of the drive. People getting high benchmarks are probably following that formula, just to get a high score...

Top 5 things that never get done:

1.


#7 cat1092

cat1092

    Bleeping Cat

  • Topic Starter

  • BC Advisor
  • 6,987 posts
  • OFFLINE
  •  
  • Gender:Male
  • Location:North Carolina, USA
  • Local time:12:00 AM

Posted 07 March 2018 - 06:46 PM

 

I do wonder, if you have a drive mostly full, with lots of small files on it, is that going to count against it in such tests?  Particularly the random access tests.  Many of those benchmarking appear to be doing so with nearly empty drives.


Yes, the notes for that test state the best results will be from an empty partition at the start of the drive. People getting high benchmarks are probably following that formula, just to get a high score...

 

 

Normally I don't quote two posts at once, however both are valid points & applies to me. :)

 

jonuk76, actually the drive isn't full, have three partitions at the first section of the drive. All are aligned, Windows 7 onwards does this by default with the latest types of drives (Advanced Format) type. Whenever creating partitions, will always later check with MiniTool Partition Wizard or AOMEI, both reports being aligned. That is important, I know this for certain. Even XP/W2K machines (SATA or IDE HDD's) benefited from the practice by not having to make two writes for every operation. 

 

Platypus, the last part of the drive (unformatted) is 963.0GB. Would this affect the test, not having a partition there? I simply left the space there as such so that as needed, can create more partitions. 

 

MDD1963, you may be onto something below. :)

 

 

Even if the results were categorized by types of drives, there are bound to be a handful of folks running a pair of whatever is fastest in RAID 0.....

 

 

I'm not 100% sure how these drives are categorized, yet the drive showing an Excellent rating must mean they're comparing to other HDD's. On the other hand, Excellent & in the bottom 10% doesn't add up. :question:

 

Joe C, when new out of box, both of these drives (have another in a XPS 8700) benched close to 200MB/sec. As stated above, both are aligned, it's a step I perform when setting up any new install, to include adding new drives, the case here. 

 

DavisMcCam, will try your suggestions & re-bench, on my W7 systems, Skype was set to Disabled via the CCleaner Startup Menu, see that I have to do manually with W10 with Tile option. Spotify, I haven't used, so have no issues with turning the Tile off. SuperAntiSpyware is Active, sometimes catching Trojans that the rest misses, yet will disable for test & report back.

 

Thanks to all for your answers, will re-bench & report back. :)

 

Cat


Performing full disc images weekly and keeping important data off of the 'C' drive as generated can be the best defence against Malware/Ransomware attacks, as well as a wide range of other issues. 


#8 cat1092

cat1092

    Bleeping Cat

  • Topic Starter

  • BC Advisor
  • 6,987 posts
  • OFFLINE
  •  
  • Gender:Male
  • Location:North Carolina, USA
  • Local time:12:00 AM

Posted 07 March 2018 - 07:23 PM

Virtually no difference from disabling the Skype, Spotify & SuperAntiSpyware apps. :(

 

http://www.userbenchmark.com/UserRun/7692222

 

Little background resources used also, as indicated by the interface (2%). My thoughts are maybe a driver or misconfiguration, just can't pin things down. The latest possible SATA drivers are installed & the HDD is connected to an Intel, not an ASMedia port. The following are the benchmarks by Crystal Disk Benchmark, sequential write looks to be way off of the money, normally has in my usage, came close to or exceeded the top number. 

 

WWgfcdy.png

 

e6RHVjm.png

 

Note that this is a clean install of W10 1709, after upgrading from W7 Pro, to 8.1 Pro w/Media Center & to all of the W10 versions, figured it was time to freshen things up. Didn't use the ASUS driver install CD as I did when building the PC, only added missing or generic drivers provided by Microsoft. Yet this performance issue was present before the clean install. 

 

Would having a partition across the last half of the drive make a difference? :)

 

Cat


Performing full disc images weekly and keeping important data off of the 'C' drive as generated can be the best defence against Malware/Ransomware attacks, as well as a wide range of other issues. 


#9 MDD1963

MDD1963

  • Members
  • 688 posts
  • OFFLINE
  •  
  • Local time:01:00 PM

Posted 08 March 2018 - 12:32 AM

 

 

I do wonder, if you have a drive mostly full, with lots of small files on it, is that going to count against it in such tests?  Particularly the random access tests.  Many of those benchmarking appear to be doing so with nearly empty drives.


Yes, the notes for that test state the best results will be from an empty partition at the start of the drive. People getting high benchmarks are probably following that formula, just to get a high score...

 

 

Normally I don't quote two posts at once, however both are valid points & applies to me. :)

 

jonuk76, actually the drive isn't full, have three partitions at the first section of the drive. All are aligned, Windows 7 onwards does this by default with the latest types of drives (Advanced Format) type. Whenever creating partitions, will always later check with MiniTool Partition Wizard or AOMEI, both reports being aligned. That is important, I know this for certain. Even XP/W2K machines (SATA or IDE HDD's) benefited from the practice by not having to make two writes for every operation. 

 

Platypus, the last part of the drive (unformatted) is 963.0GB. Would this affect the test, not having a partition there? I simply left the space there as such so that as needed, can create more partitions. 

 

MDD1963, you may be onto something below. :)

 

 

Even if the results were categorized by types of drives, there are bound to be a handful of folks running a pair of whatever is fastest in RAID 0.....

 

 

I'm not 100% sure how these drives are categorized, yet the drive showing an Excellent rating must mean they're comparing to other HDD's. On the other hand, Excellent & in the bottom 10% doesn't add up. :question:

 

J

 

Cat

 

Very good point!


Asus Z270A Prime/7700K/32 GB DDR4-3200/GTX1060


#10 cat1092

cat1092

    Bleeping Cat

  • Topic Starter

  • BC Advisor
  • 6,987 posts
  • OFFLINE
  •  
  • Gender:Male
  • Location:North Carolina, USA
  • Local time:12:00 AM

Posted 08 March 2018 - 05:35 AM

MDD1963, yes a good point and beginning to wonder if I should swap drives. :(

 

The frown is up there because these are (kind of like the SATA-2 WD RE4 was to the Caviar Black) now improved, faster than ever & still just $10 more than today's WD Black (SATA-3) when on promo. The Disk tests shows the drive is faster than the RE4 or today's Caviar Black. I mean, with a 128MB cache & running at 7200 rpm, it should be. On the other hand, the WD Gold (any size) are prosumer drives, really designed for RAID and/or Datacenter usage. So unlike the RE4 once done for me (after running the wdidle3 utility), has became an expensive liability, preventing me from reaching UFO status once by a single point in the Desktop rating when new. At $129.99 each on promo, which still isn't a bad deal for a HDD with a 5 year warranty, makes me wonder 'what if' I'd saved a total of $20 on both & purchased the same size WD Blacks instead. The main difference is half the cache. 

 

Well, with enough (or excess) cash invested in storage, and a couple of fast 1TB Seagate ST1000DM003 HDD's on hand, will have to swap & at least make the attempt to break the barrier. That's what was installed before the WD Gold was, there's been other system components upgraded (notably RAM) and now have the CPU tuned to near perfection (running at 4.4GHz by adjusting the clocks for all cores by 44). So my belief is that if I can find a drive onhand that will test in the top 85-90%, that should put me over the hump. After all, the drive test is one of those that falls under the Desktop rating. 

 

While I'd prefer that UserBenchmark would allow me to stick a 2.5" SSD in there for the all three UFO ratings, that's not how their benchmark system works, points are knocked off if there's not a spinner.........what a bummer! :o

 

Am I in the year 2018 or are they 5 years backwards in time? :question:

 

Sure seems the latter happens to their case. The site does have great in-depth reviews of some great & (in particular) the latest CPU's, all ranked from high to rock bottom. Great source of knowing where one stands or purchase consideration. The only thing I can't seem to figure out is that the 2TB WD Gold is ranked #1 out of 984 HDD's & even the Poor range is 85%. Yet I can barely break 10% most of the time. :(

 

http://hdd.userbenchmark.com/SpeedTest/206650/WDC-WD2005FBYZ-01YCBB1

 

What I believe I'll do is stick the drive in another SATA-3 PC & see if there's a difference, don't know what else to do. The WD extended test came back with no issues, Maybe there's some (invisible) software dragging down performance or it's W10. However, one's also docked by not having the latest OS installed, so it's dang if I do & dang if I don't. 

 

Could older drivers be the answer? Under normal circumstances, I'd not think so, Intel MB's tends to fare well with the latest drivers. I simply cannot see this drive scoring so low on the test, and see in real action, there's two times when it hits the 4K speed test, that the test shows 'waiting, slow I/O load'. (or words to that effect) That's the only giveaway that something's not quite right. 

 

What affects the I/O load, speed or whatever the term means? The system drive (512GB Samsung 950 PRO NVMe SSD) passes with flying colors (276%,around the top 2 percentile), plugged directly into the 2nd PCIe slot via adapter, wanted to keep it off the hot MB, plus that M.2 connection is Gen2, so half the speed if were there. There's otherwise only two drives, one optical & the drive being discussed in this Topic, so the SATA controllers being overloaded isn't an issue. The only other option I may have, would be to install a SATA-3 card to the final GPU lane, which is the single 2.0 one (the other two are PCIe 3.0). Only thing being, it's a low cost IOCrest card & would have to find the tall bracket, don't see how it could best the onboard Intel SATA controller (two ports of which are SATA-Express). Have tried all Intel ports, plus the two ASMedia ones to no avail. 

 

Will keep shoveling! :workout:

 

Cat


Performing full disc images weekly and keeping important data off of the 'C' drive as generated can be the best defence against Malware/Ransomware attacks, as well as a wide range of other issues. 


#11 Platypus

Platypus

  • Moderator
  • 13,700 posts
  • OFFLINE
  •  
  • Gender:Male
  • Location:Australia
  • Local time:02:00 PM

Posted 08 March 2018 - 06:06 AM

Platypus, the last part of the drive (unformatted) is 963.0GB. Would this affect the test, not having a partition there? I simply left the space there as such so that as needed, can create more partitions.


I can't see that affecting the test, it can only test the partition that's there, and it is in the faster but not necessarily the fastest portion of the drive. There could be a noticeable increase from having an empty partition for the test, at the very start of the drive where the DTR will be the highest.

 

But there does seem to be something weird with the read and write parameters pretty much reversed to what you'd expect, and what you see in the Crystal report.


Top 5 things that never get done:

1.


#12 Joe C

Joe C

  • Members
  • 775 posts
  • OFFLINE
  •  
  • Local time:12:00 AM

Posted 08 March 2018 - 06:42 AM

Looks like the write takes longer than the read time, (read alway's faster)  which is normal for any type of drive?



#13 cat1092

cat1092

    Bleeping Cat

  • Topic Starter

  • BC Advisor
  • 6,987 posts
  • OFFLINE
  •  
  • Gender:Male
  • Location:North Carolina, USA
  • Local time:12:00 AM

Posted 08 March 2018 - 07:08 AM

Looks like the write takes longer than the read time, (read alway's faster)  which is normal for any type of drive?

 

Yes that's right, read is usually faster near 100% of the time. :)

 

What kind of shown as a red flag to me however, was the Crystal Disk Benchmark above. Notice how the Sequential test (3rd set of numbers) on the read actually best the top number. Well, the sequential write is also supposed to be close to the top number, maybe not surpass it, yet should at least be close, regardless of HDD. That must be what's killing the performance. Very poor sequential write speed. Below is a snapshot of the drive when first installed (current speed in Post #8) & there was much more software on the system at the time, in fact, was one upgrade upon another. See how the sequential write speed exceeds the top number? Today, I cannot replicate that speed. :(

 

WL5rHuR.png

 

This is the problem at hand, IMO. 

 

Cat


Edited by cat1092, 08 March 2018 - 07:16 AM.

Performing full disc images weekly and keeping important data off of the 'C' drive as generated can be the best defence against Malware/Ransomware attacks, as well as a wide range of other issues. 


#14 DavisMcCarn

DavisMcCarn

  • Members
  • 732 posts
  • OFFLINE
  •  
  • Gender:Male
  • Local time:12:00 AM

Posted 08 March 2018 - 08:03 AM

OK, lets backup a little bit here.

A traditional rotating hard disk drive's performance is governed by several factors with the lowest hardware level being the areal density (bits per centimeter) and the rotational speed of the platters.  Modern drives change the areal density as the tracks get smaller (closer to the center) in order to improve the reliability of read/write operations.  As a result, the fastest read/write areas are at the beginning (outer tracks) on the media.

Additionally, one of the biggest losses of actual data transfer speed happens when seeking is performed, meaning moving the heads to another track.  Most drives need about 12 milliseconds to move the heads and, using your own numbers, the drive could have supplied another 2.268 million characters of data while it was moving the heads (seeking).

On poster earlier stated that optimal numbers came from benchmarking an empty partition at the beginning of the drive which most certainly jives with the physics of how drive's operate.  As one moves towards the center of the media, the data transfer rate diminishes and seeking to avoid files already on the drive slows it down even more.

Caching (the 128 or 256 megabytes you referenced) is only a factor in that the drive will cache the most recently accessed or most accessed areas of the media so it won't have to read the media surface in order to supply that data to the host PC and it's efficiency is determined by how the drive is being used and the algorithms written by the manufacturer for how to  use that cache.  If you are accessing larger files, the cache will have less impact because they do not fit into the cache so the drive must go and actually read the media.  Folders or smaller files will improve the cache's efficiency.

SSD's overcome these physical issues as the data transfer is not governed by the areal density problem and a fixed rotational speed.  They also typically can reference another area of storage with a one millisecond delay rather than twelve.

So, if you want to get near the top of the benchmark numbers for that WD, delete all of the partitions (after backing up, of course), then create a 100GB partition at the beginning of the drive, format it but leave it empty, and then rerun the benchmarks.

As a note, we considered 11 MBps to be very fast in the mid-1990's.


Computer dinosaur, servicing PC's since 1976

#15 Joe C

Joe C

  • Members
  • 775 posts
  • OFFLINE
  •  
  • Local time:12:00 AM

Posted 08 March 2018 - 10:23 AM

Have you defragged this platter drive recently?

I have a couple 3tb drives and a 1tb drive in my system and I notice a slowdown after a while to access files on the platter drives. I do not defrag the ssd but the platter drives will access much better after a defrag. I use a third party software called Vopt to defrag my platter drives






0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users