OCZ Vertex 3 Max IOPS - 75,000 IOPS???

[JohnnyLucky]

Yesterday morning OCZ released the OCZ Vertex 3 Max IOPS solid state drive which boosts maximum IOPS to 75,000.

In an official OCZ press release, OCZ CEO Ryan Peterson is quoted as saying:

"Vertex 3 Max IOPS drives increase random write performance, and are the ideal storage solution for applications that require high aggregate workloads and increased IO throughput."

What would a typical home user, a gamer, or an enthusiast be doing that would require such an enormous IO throughput capability? Is there software or a game that could make use of it? I have not read about about anything like that for home use or gaming.

There are two ways to measure IOPs. The first is the typical benchmark which measure how many I/O's a solid state drive can process in one second. That is the benchmark we almost always see. The second way involves measuring the actual IOPS while doing something like playing a game or editing an image in Photoshop. That is the measurement that we typically do not see. Luckily there are a few reviews that provided the second measurement. In those benchmarks the IOPS rarely exceed 4,000.

I've never seen an article that reported unusually high IOPS for typical home use or while gaming. Anand over at AnandTech suggested 20,000 IOPS is the practical upper limit for ssd's.

What do people do on their home computers or while playing games that require such high capabilities?

FTR - Over on the business enterprise side there are SSD's that are capable of 1 million IOPS but they are highly specialized ssd's for highly specialized applications. They are not suitable for home use or gaming. Don't even ask about the price.

 

[JohnnyLucky]

No comments? I'm surprised. I was hoping someone might have an answer or a reference to an article.

 

[tecmo34]

I read an article but can't find the link but it is related to Facebook and zynga games. They require MAX IOPS... :lol:

Hey, you got an answer... just not an accurate one

Probably a more suitable answer would be utilization of VDI (Virtual Desktop Infrastructure).

http://virtualizationjedi.com/2010/10/31/finding-a-better-way-to-estimate-iops-for-vdi/

 

[JohnnyLucky]

Thanks!

Interesting article and usage model - 77,000 boot IOPS, 17,500 steady state IOPS, and 28,800 peak IOPS for a storage array serving a whopping 3,500 business desktop pc's operating at a variety of loads. I only have one pc so I can't test it. :lol:

 

[ricno]

What do people do on their home computers or while playing games that require such high capabilities?

I ordered my 240 GB Vertex 3 some time ago and it arrived yesterday and will mount it tonight. My reason for it is that I mostly use my home computer (besides web browsing) to run virtual machines. With a fast CPU (core i7) and lots of RAM (16 GB) it fits many virtual machines at the same time and I do a lot of testing, learning and experimenting with server operating systems this way (work related).

In some ways a mechanical disk limits the performance when running multiple VMs which does a lot of random IOs, so a SSD should be perfect for this. This Vertex disk will not be used for the main operating system but as a datadisk only.

 

[flong]

Hi, I think that most people buying the SSDs are after the faster overall performance. And when you think about it, only about $100.00 separates the cutting edge SSDs from those that are going extinct. Nearly every review on SSDs state that they are amazingly fast and save boot times, load times, transfer times, download times, etc. I for one would pay an extra $200 over an HDD just for that alone. I cannot tell you how much I hate waiting on the computer to catch up with me.

I do some CAD work, some photo work (not much). I watch HD video and HDTV - I have a TV card and watch OTA HDTV. I don't really game. I do surf the web a lot and watch HULU and other streaming services.

Many times when I am working I will have five or six applications open while I am listening to the radio over the internet and I can tell you that my computer has a hard time keeping up.

I have never owned an SSD but as I said, If it improves my overall computer experience speed-wise at all, it will be easily worth the extra $200. Time is money.

As far as there being one specific software or game, game that requires the super speed of an SSD compared to an HDD - probably only CAD, video editing and maybe photo editing would be in the realm of normal users. CAD does not necessarily require it - I have an XP dual-core desktop computer and my CAD program works reasonably well on it. But some CAD programs use a ton of resources (mine doesn't use that much). The computer I am building will be over-kill but it will also be wicked fast and that is what is important to me.

 

[groberts101]

ricno. for that usage scenario you may want to give Fancycache by Romex a look as it will reduce all that little redundant stuff from being written as heavily. You can defer the writes to group them in small batches which saves the drive from dealing with so many partial blocks and the subsequent juggling that it requires. It has been said to speed up those specific data sets by others running VM's as well.

http://www.romexsoftware.com/en-us/fancy-cache/

 

[ricno]

ricno. for that usage scenario you may want to give Fancycache by Romex a look

Thanks, I will take a look at that after I have the SSD installed and tryed out a bit.

 

[JohnnyLucky]

ricno - understood. I've read some interesting articles about VM's. One of the things I've seen mentioned is the need to conduct actual experiments with ssd's to find out what works best for a particular scenarios. It's also good read that somone is experimenting. That's one of the fun things about personal computers.

flong - I understand. When it comes to ssd's or any other new component people have really high expectations. The problem is people might be misled by synthetic benchmarks. Home users, gamers, and enthusiasts might be led to believe that really high IOPS numbers are required for optimal performance. At least that is what I think is happening. It might be time for a reality check. Home users, gamers, and enthusists might be surprised to learn that a peak or burst of 4,000 IOPS is all that is actually needed for optimal performance when multitasking, gaming, or during a heavy workload session.

 

[flong]

ricno - understood. I've read some interesting articles about VM's. One of the things I've seen mentioned is the need to conduct actual experiments with ssd's to find out what works best for a particular scenarios. It's also good read that somone is experimenting. That's one of the fun things about personal computers.

flong - I understand. When it comes to ssd's or any other new component people have really high expectations. The problem is people might be misled by synthetic benchmarks. Home users, gamers, and enthusiasts might be led to believe that really high IOPS numbers are required for optimal performance. At least that is what I think is happening. It might be time for a reality check. Home users, gamers, and enthusists might be surprised to learn that a peak or burst of 4,000 IOPS is all that is actually needed for optimal performance when multitasking, gaming, or during a heavy workload session.

I believe you and it is an excellent point. I for one do not really understand IOPS relation to overall computer speed in numbers as you do. As just a desktop user I am going by the reviews and first-hand owner accounts of the increased speed of SSDs for the overall experience.

For example, I spent over $300.00 on the ATI 5850 a couple of years ago when I owned an I-7 920 system. I upgraded from a cheap Nvidia card to the 5850. A bunch of people said that the card that I had was sufficient and wouldn't make much of a difference, but I cannot tell you how much a difference it made in the OVERALL speed and quality of my system. I don't really understand how it did it but it was an amazing card that produced crystal clear HD, great colors, quick speeds, great HDTV and more. The difference was night and day.

In the same way, I think that the SSD will improve the user experience based on the statements of those who have upgraded to SSDs from HDDs. I get what you are saying that most users other than heavy gamers don't really need an SSD to run their applications (and even gamers don't really have to have an SSD) but it is the quality of the experience I am looking to improve. I hate waiting on the computer. While I am not a programmer, I am a very savvy user and I am an expert with lots of different software. When I multitask, I even taxed the 920 system with 16 GB of RAM I had and it was a very fast computer. I am hoping that an SSD will cut down on wait times for different applications.

 

[JohnnyLucky]

Today I crossed back over to the mainstream/business side. I've found some interesting news articles about IOPS.

SSD's have been around for a good ten years. They are not new. They were installed in servers used by businesses and they were very very expensive. The three main IOPS utilities were used to measure IOPS requirements for the data drive arrays in servers. For example, a company that had a server and 1,000 desktop pc's would set up a data drive array for 100 of the pc's and measure the IOPS during actual real world use. The results would be used to determine data drive array and IOPS requirements for the server that fed all 1,000 pc's. The test procedure is missing in reviews of consumer ssd's. With three exceptions, I have not seen any similar results for an ssd installed in a personal desktop pc or gaming rig. For example, I still have not found a review or report that measured the actual IOPS during a Starcraft II, single player, session.

Beginning in 2009 articles and white papers started appearing indicating that the IOPS benchmarks were being abused. Company advertising aimed at the emerging consumer market focused on maximum IOPS instead of actual requirements for maximum optimal performamce. The actual requirements would depend on the hardware configuration, the software application, and how they were used.

The situation reminds of power requirements for pc's. Gamers and enthusiasts tended to purchase high wattage power supplies even though reviews that measured actual power usage clearly indicated less power was required than individuals thought.

 

[flong]

Today I crossed back over to the mainstream/business side. I've found some interesting news articles about IOPS.

SSD's have been around for a good ten years. They are not new. They were installed in servers used by businesses and they were very very expensive. The three main IOPS utilities were used to measure IOPS requirements for the data drive arrays in servers. For example, a company that had a server and 1,000 desktop pc's would set up a data drive array for 100 of the pc's and measure the IOPS during actual real world use. The results would be used to determine data drive array and IOPS requirements for the server that fed all 1,000 pc's. The test procedure is missing in reviews of consumer ssd's. With three exceptions, I have not seen any similar results for an ssd installed in a personal desktop pc or gaming rig. For example, I still have not found a review or report that measured the actual IOPS during a Starcraft II, single player, session.

Beginning in 2009 articles and white papers started appearing indicating that the IOPS benchmarks were being abused. Company advertising aimed at the emerging consumer market focused on maximum IOPS instead of actual requirements for maximum optimal performamce. The actual requirements would depend on the hardware configuration, the software application, and how they were used.

The situation reminds of power requirements for pc's. Gamers and enthusiasts tended to purchase high wattage power supplies even though reviews that measured actual power usage clearly indicated less power was required than individuals thought.

You probably already know this but Anand-tech has IOPS tests in their SSD reviews. I don't really understand them but they are there.

 

[JohnnyLucky]

AnandTech is one of the sources of information that made me think twice about IOPS benchmarks. Anand measures average IOPS for a heavy workload, a light workload, and a gaming workload. The results are included on page 8 in this morning's review of the new OWC ssd:

http://www.anandtech.com/show/4315/owc-mercury-extreme-pro-6g-ssd-review-120gb/8

In the text Anand explains exactly what software, applications, and utilities were used and what they were doing when the measurements were taken for the heavy and light workloads. If a game was used to measure the gaming workload it was not mentioned.

As you can see in the results the average IOPS were on the low side which may surprise gamers and enthusiasts. Although not mentioned, I think the burst or peak IOPS would probably be somewhere around 4,000 IOPS for the top performing ssd's.

The big surprise was the gaming workload. Two thirds of the ssd's made up a very tight cluster. The measurements were between 309 IOPS and 325 IOPS. Remember, this is the average IOPS during use, not the burst or peak IOPS. I'll see if I can locate the two reviews that included burst and peak.

 

[metawin]

The max IOPS edtion uses a better grade NAND

 

[flong]

The max IOPS edtion uses a better grade NAND

I am hearing the th 120 GB IOPS is the best 120GB SSD but I have not read any reviews yet. Do you know what nand they IOPS is using?

 

[JohnnyLucky]

I am hearing the th 120 GB IOPS is the best 120GB SSD but I have not read any reviews yet. Do you know what nand they IOPS is using?

it's a good thing I have a database I compiled.

Toshiba TH58TAG7D2FBAS9 16GB, 34nm, Toggle-NAND Flash same as the Vertex 3 Pro.


OCZ Vertex 3 Max IOPS Reviews:

http://benchmarkreviews.com/index.php?option=com_content&task=view&id=748&Itemid=60

http://thessdreview.com/our-reviews/ocz-vertex-3-240gb-max-iops-review/

http://www.tweaktown.com/reviews/4110/ocz_technology_vertex_3_max_iops_240gb_ssd_review/index.html

http://www.kitguru.net/components/ssd-drives/zardon/ocz-vertex-3-240gb-max-iops-ssd-review/

http://www.hardwareheaven.com/reviews/1179/pg1/ocz-vertex-3-max-iops-240gb-ssd-review-introduction.html


The Vertex 3 Pro was a business enterprise prototype with either a SandForce SF-2582 or SF-2682 controller. It was not designed for gamers and enthusiasts. OCZ announced it would be available last March but I don't think the ssd was ever released.


OCZ Vertex 3 Pro (prototype) Reviews:

http://www.legitreviews.com/article/1547/1/

http://www.guru3d.com/article/ocz-vertex-3-pro-review/

http://www.pcper.com/reviews/Storage/OCZ-Technology-Vertex-3-Pro-SandForce-2582-200GB-SATA-6G-SSD-Review?aid=1078&type=expert&pid=10

 

[flong]

it's a good thing I have a database I compiled.

Toshiba TH58TAG7D2FBAS9 16GB, 34nm, Toggle-NAND Flash same as the Vertex 3 Pro


Reviews:

http://benchmarkreviews.com/index.php?option=com_content&task=view&id=748&Itemid=60

http://thessdreview.com/our-reviews/ocz-vertex-3-240gb-max-iops-review/

http://www.tweaktown.com/reviews/4110/ocz_technology_vertex_3_max_iops_240gb_ssd_review/index.html

http://www.kitguru.net/components/ssd-drives/zardon/ocz-vertex-3-240gb-max-iops-ssd-review/

http://www.hardwareheaven.com/reviews/1179/pg1/ocz-vertex-3-max-iops-240gb-ssd-review-introduction.html

Did I understand you right that the Vertex 3 has gone back to 34 nm NAND? BTW you are efficient as well as beautiful lol.

 

[JohnnyLucky]

The technical reviews indicate and show photos of the Toshiba 34nm NAND flash

 

[palladin9479]

Consumer level activity (as currently expressed) shouldn't have anywhere near the requirements of Enterprise equipment. Any commodity SSD should provide more then enough I/Os. Processing a database that is being accessed by 1000+ computers is a whole helluva lot different then playing Metro 2033, compiling code or processing video. IOPS is just being thrown around as a marketing word, its pretty meaningless for everyday use.

 

[JohnnyLucky]

palladin9479 - I absolutely, positively, 100% agree. What I don't understand is why the vast majority of authors who publish technical reviews at web sites that cater to gamers and enthusiasts fail to explain that.

 

[palladin9479]

Because its another number that can be pointed to to say "X produce performs 10% more then Y product". Current SSD's are getting pretty dead even on performance levels, people need things to distinguish them apart.

 

[flong]

Consumer level activity (as currently expressed) shouldn't have anywhere near the requirements of Enterprise equipment. Any commodity SSD should provide more then enough I/Os. Processing a database that is being accessed by 1000+ computers is a whole helluva lot different then playing Metro 2033, compiling code or processing video. IOPS is just being thrown around as a marketing word, its pretty meaningless for everyday use.

That being said, the IOPs version of the Vertex 3 240GB is faster than the standard version judging from the reviews - but not a lot.

 

[JohnnyLucky]

flong - What would you do with all those IOPS? Do you have some sort of software application that would require such a high number of IOPS for optimal performance?

I am not aware of any consumer grade software applications or games that could make use of 75,000 IOPS.

 

[flong]

Hi JL, it is good to hear from you again . As I have mentioned in other posts, I am not so concerned about IOPS or all the other descriptions of an SSD, I am concerned about its overall performance in the real world. The IOPS is just one thing to look at, but what I really look at are the benchmarks in the professional reviews.

So for example, if in the benchmarks of a professional review the 120GB IOPS V3 is measurably faster than the standard 120GB V3 running Windows, MS Office, Windows MC, CAD, video software (VLC), etc. that is what I am looking for - it is . They can call an SSD whatever they want, I am not an electrical engineer and so I do not profess to understand all of the inner workings of an SSD. What I do care about is how it performs in the real world.

It can be tricky to sift through the different labels to find the best SSD for a particular situation. For example the Agility 3 120 GB is as fast or nearly as fast as the V3 but it is $100 cheaper. Tom's Hardware didn't choose it as a best buy in their recent review. They did mention the 240GB Agility. So even the pros are a little confused right now ha, ha.

 

[adampower]

For example the Agility 3 120 GB is as fast or nearly as fast as the V3 but it is $100 cheaper. Tom's Hardware didn't choose it as a best buy in their recent review. They did mention the 240GB Agility. So even the pros are a little confused right now ha, ha.

Exactly, even the pros are confused. Since NAND density has doubled I have doubled my minimum 'performance' drive expectations. To clarify, the 60GB Vert. 2 did not perform as well as the 120gb vert 2. But performance almost flattened out above 120. One can say that 64 gb of 34 nm NAND was not quite enough to saturate the sandforce 1200 and its 8 channels. So, now we require 256 gb of 25nm NAND in order to make 8 channels of sandforce sing. At least that's my theory.

However, the 120gb vert 3 works great. Not to mention the 120gb m4 by the way. Still, I'm holding out for the price of 240 gb ssds to come in range. Oddly enough it's not that I need the space on my ssd. Its more the performance.

Johnny raises a great question. When is enough enough? 10k iops? 50k? 1 million iops? Even Tom's ssd guide seems to answer the question with gb/$. Even an intel drive is great when the price is right!