I modelli OCZ RevoDrive si presentano nella veste di scheda PCI-Express, dotati di connettore fisico x4. Siamo dunque di fronte a una soluzione anomala, che non passa per i classici connettori Serial ATA. OCZ ha realizzato queste unità per gli utenti più esigenti in ambito home, come i gamers che non vogliono compromessi, oppure per l'utilizzo in ambito server .
Concettualmente we are faced with something that resembles the old RAID controller , although in this case are on the card also supports physical storage. Looking deeper into the logic of management, necessary to make assumptions. The usual 2.5-inch SATA drives are actually implemented by configuring "in RAID 0 individual chip memory, with a number of different lines depending on the model. To manage it all provides a controller, which in turn can communicate with the chip that will handle the final flow to and from the physical cable SATA.
RevoDrive In the case of the situation is complicated. Besides the already named link "hidden" in RAID 0 chip memory, there is a further aggregation of different memory stack, which are present in the model RevoDrive number 2, a 4 instead RevoDrive X2 models. To illustrate, there are two models RevoDrive SandForce 1200 series controllers, each pledged to hold off two different memory blocks. These two stack converge into a single chain, then RAID 0, managed by another house Silicon Image controller.
 4 Detail of the controller 1200 on SandForce RevoDrive RevoDrive X2 X2 With the situation comes up again, this time, however, stack with 4 controllers and 4, always SandForce 1200. This explains why these units achieve stated performance far superior to traditional 2.5- inches. A veritable cascade of chains converge on a single RAID 0 controller, which then smisterà the I / O PCI-Express, chose not to have bandwidth problems. Given the performances, which exceed the 700MB per second read, declared, it is clear that not only the Serial ATA 3Gbps was inadequate, but also to 6Gbps would have been a bottleneck.
OCZ RevoDrive and RevoDrive X2 - continued  | The two models of different generations are present in a list with great choice in terms of capacity. We try to figure out how to make up the range of OCZ RevoDrive, analyzing the common traits and differences.
| | RevoDrive | RevoDrive X2 | | Cuts Available | 50GB to 480GB 100GB to 960GB | | | interface | PCI-Express x4 | | Configuration | RAID 0 to RAID 0 | two channels to four channels | | Controller | SandForce 2 x 1200 4x | SandForce 1200 | | MBTF | 2,000,000 hours | | consumption | Idle 3W, 8W 4.3W active | Idle, 8.3W active | As you can see RevoDrive X2 is available in denominations of exactly twice as RevoDrive, suggests that a simple doubling of the memory banks and controller to achieve RevoDrive X2 starting from the previous project. Hence RevoDrive that is offered in denominations from 50GB to 480GB, while RevoDrive X2 reaches well in 960GB, 100GB starting.
The interface is common, PCI-Express x4, as well as the high value of MTBF 2 million hours. Similar to the consumption, and are the doubling already mentioned in terms of adoption of on-board controller SandForce. The native configuration is guaranteed by the Silicon Image controller RAID 0 two-way models RevoDrive, rising to four for RevoDrive x4. It can be seen in the image as OCZ has in fact structured RevoDrive X2 dual card, against the single RevoDrive.
| | | Performance declared | | | | Scripture Reading | | | RevoDrive | Models from 50GB to 80GB | up to 450MB / s up to 540MB | / s | | | Models 100GB to 160GB | up to 480MB / s | up to 540MB / s | | RevoDrive X2 | Models 100GB to 160GB up to | 690MB / s up | to 740MB / s | | | Models from 240GB to 960GB | up to 720MB / s | up to 740MB / s | These performances of the two models are very high, but justified by the discourse done on the previous page. Interesting to note, however, there are two groups within each family, with higher performance models for a given capacity up. For RevoDrive 50GB to 80GB models are credited to 450MB / s write and 540MB / s maximum read, while the cutting unit above the maximum write speed goes up to 480MB / s. For
RevoDrive X2 values \u200b\u200bare even higher: 100GB to 160GB models are credited to 690MB / s maximum write, with a value of up to read well-740MB / s. From 240GB up the write speed goes up to 720MB / s. The models in the test belong to both groups performing, because in our laboratories have come RevoDrive RevoDrive X2 120GB and 420GB.
The equipment is rather limited, since it includes the storage unit, a well-designed packaging and a package for quick installation, with very minor suggestions.
Considering that the users are going to buy this kind of unity is understandable not to find a book in every way; remains questionable choice not to include the drivers necessary for use in environments Microsoft Windows, available on the OCZ site . |
SSD Mach Xtreme MX-Technology 120GB DS | | We decided to combine the two models also Revodrive OCZ SSD Mach Extreme Technology, MX-120GB DS. The decision was soon justified: This unit is equipped with a controller SandForce SF-1200, which is why we can define a scenario in which they are present in the same controller numbers one, two and four, and observe the scalar performance.
The form factor of the model MX-DS is 2.5-inch 120GB, as most of the SSD market. The company provides in the package, including a bracket for mounting in a bay 3.5-inch screws. Aesthetically, nothing to report, other than the street-art style graphics of the label. An aluminum chassis houses the usual components, adapting to the standards as regards the overall dimensions.
Deprived of small cap, fixed by four screws, the disc shows the normal components on a single card that characterizes the production 2.5-inch SSD. You can see the memory chip (MLC), the controller and the small mail to the SATA connector. The interface chosen by Mach Technology is the Extreme SATA 3Gbps, suits guaranteed performance from the controller used.
Here the controller, then. The unit that equips SandForce SF-1200 Mach Technology Extreme MX-DS ago the company declared values \u200b\u200bof transfer rate up to 280MB per second read and 265MB per second in writing, and maximum values \u200b\u200bof up to 30,000 IOPS ( here find the official page). Then we enter the heart of the comparative analysis of the methodology of testing. |
The methodology for testing | |  Test Sample inconsistent presence in the Silicon Image controller One of the most difficult to deal with when it comes to solid-state drives is the procedure for testing. Things get complicated when SSD to be affected are included are the Revodrive of our benchmarking. The presence of multiple controllers in cascade can send in crisis some tools normally used in our analysis. E 'success with HD Tune Pro, we dropped in this analysis because of the inconsistency of the data (graphs very jagged and with values \u200b\u200bin MB / s never over 280MB / s, even with the most powerful Revodrive). We have therefore chosen to use a suite for the occasion, which allowed us to obtain the desired results. These software used for our analysis:
1. AS SSD - speed read and write speed - transfer rate, data copy
2. ACT 2.46 - read speed and write speed - transfer rate
3. Iometer Test - Transfer rate and IOPS
Different scenarios examined by Iometer. We analyzed the test unit into sequential read with packages of varying sizes: 512B, 4KB, 128KB, 152KB, 1MB and 2MB, while we have chosen to write the worst case scenario, a 4K random writes. For all tests were also tested with different behaviors Queue length, true ov 1, 8, 32. For the analysis we examined more IOPS random write the 4K with Queue length of 1, 8 and 32 in order to undermine the greatest possible unity, choosing instead to read sequential packets to 16K to get the maximum in terms of IOPS. Here is the location used.
- Processor: Intel Core i7 920, clocked with 3.2 GHz Turbo Mode active
- RAM: 3x2 Gbytes Corsair TR3X6G1600C8D; clocked at 1,600 MHz with timings of 8-8-8-24;
- Hard disk : Intel X-25M SSD 80GB SATA 3Gbps
- Motherboard: Gigabyte GA-X58A-UD7, chipset Intel X58;
- System OS: Windows 7 Ultimate 64bit, with all the latest fixes issued by Microsoft;
- Power: 1100 Watt Topower. | AS SSD - Read and write | | start the analysis with AS SSD performance benchmarks, analyzing the behavior in sequential read and write, in addition to the scenarios that take into consideration the packets from 4K and 4K in 64 threads.
 Mach Xtreme MX-Technology 120GB DS
 120GB OCZ Revodrive  Revodrive X2 240GB OCZ
We have summarized for ease of reference the results in a graph, it appears to the eye can see the comparison between the units of analysis.
sequential read and write have proved in line with expectations. The MX-DS reached in this test the 170MB per second in reading, writing and stops just under 90MB per second. Results are obviously different units OCZ Revodrive, with the dual controller model next to 380MB per second read and write almost 190MB per second. Important results Revodrive X2, we are about 660MB per second read and 360MB per second in writing, confirming a certain linearity in the scale of benefits based on their configurations and controllers used in RAID 0.
By reading 4K random, single and 64 threads, things change a lot. Immediately apparent, and clearly, as the units OCZ Revodrive are much more than a whim for fans We are facing units can handle heavy workloads, where a single SSD, but still very powerful, it shows the side. Soon only on test 4K single thread, as it allows the unit to operate at best at least in terms of transfer rates. Unlike the discourse on the IOPS, we'll see below.
find in writing the scenario as above, although with a maximum in a single test thread above. This is a very challenging scenario for the storage subsystem, as shown by values \u200b\u200bvery distant from the maximum stated.
AS SSD - Text File | | integrated in AS SSD Benchmarks are a convenient tool that allows you to measure time in seconds required to move three data packets. The first is an ISO image, with sequential data so far, followed by Program (several files of different sizes) and Game, which arises in the middle like scenario. All three packages have sizes ranging from 1GB to 1.4 GB. Here are the results.
 Mach Xtreme MX-Technology 120GB DS
 120GB OCZ Revodrive  X2 240GB OCZ Revodrive Again, we collected the results in a chart obtained, while above the detail of individual benchmarks for individual units.
We find a measure of proportionality between lap times and configurations of the individual units. OCZ Revodrive X2 takes about 5 seconds to copy the game package, and they are almost double in OCZ Revodrive and more than triple the MX-DS. Similar situations for other scenarios. Worthy of note are only 3.6 seconds to copy more than 1GB of ISO image with the OCZ units Revodrive X2. |
ACT Benchmark - sequential read and write | | keep the pattern seen earlier, showing screenshots of each test and then comment on them once they are included in the graphs.
 Mach Xtreme MX-Technology 120GB DS
 120GB OCZ Revodrive
 OCZ 240GB X2 Revodrive
The line chart can highlight the behavior of sequential read-write unit tests, at least for the results obtained with ACT Benchmark. We have shown for convenience of reading only the packets from 4K, 128K, 512K and 2M, which allow you to analyze the performance scale with increasing heaviness of work required at the storage subsystem.
emerges as the best unit from OCZ Revodrive you get with the exchange of packets that exceed 512K, although the performance record below this size are still very high. The DS-MX SSDs ago definitely "wall" just reached its limit already to 128K, with a constant rate with increasing packet size.
In sequential write, the situation changes a little, with the OCZ units Revodrive X2 excels as expected. Interestingly, this limit is 128K with packages for MX-DS for OCZ Revodrive, beyond which the units referred to settle on their best, in this particular test. |
ACT Benchmark - Random | |  Mach Xtreme MX-Technology 120GB DS
 Revodrive 120GB OCZ
 X2 240GB OCZ Revodrive
average lower values \u200b\u200bfor the causal reading of the data, although all values \u200b\u200bobtained to maintain the proportionality already seen in other tests. The observations made above regarding the performance related to the size of packets in this round of testing.
In writing random things even worse, but we must remember that the limit in a scenario where writes occur randomly forced, remote possibility of all scenarios in "real-life". Note that the maximum in terms of transfer rate will be drastically reduced, even for the most powerful units of the lot.
| |
|
Iometer - Sequential read | | We chose to analyze the behavior in the opposite scenario using Iometer performance tests. The first looks at the sequential read, then set 100% 100% sequential reads. This test required a fairly complex pattern of use, as we asked Iometer to run with different packages of 512B, 4K, 16K, 128K, 512K, 1M and 2M, each with different queue length: 1, 2, 4, 8, 16 and 32. Each disk was analyzed sequential read and then subjected to measurements in 42 different scenarios, of which we represent a summary (512B packets, 4K, 128K, 512K, 1M and 2M queue length of 1, 8 and 32).
The sequential reading observed with Iometer test repeated when it is confirmed, the previous pages using other tools. The values \u200b\u200brise with the size of packets analyzed, combined with action to expose the limits of individual units in terms of MB per second maximum. Note that the maximum values \u200b\u200bof the two models OCZ do not differ much then data declared by the significant, while it seems far more the MX-DS, even stopping to consider the values \u200b\u200bof respect to the type of unit, taking into account the test with a queue length.
Increasing the queue length of 8, we notice that is required the maximum disk subsystem with smaller packages. In this case the results dell'OCZ Revodrive X2 even exceed those stated, although it should be considered that, just as with the completely random writes to 4K, this scenario is certainly not a reflection of normal use of the units.
somewhat similar behavior with queue length 32, since the unit were already put under stress with the previous test. Bedding maximum sequential read performance around 800MB per second for OCZ Revodrive X2, 510MB per second and Revodrive OCZ 203MB per second for MX-Xtreme DS Mach Technology.
Iometer - IOPS | | The number of IOPS, trumpeted with great show of optimism on the part of many companies that sell SSD, is perhaps one of the most complex to measure. The reason is inherent in the nature of the acronym itself, I nput / O Output perations P er S ECOND . The maximum declared very often is exaggerated. Not only that: according to the scenarios analyzed, this can be drastically reduced, so as to be almost misleading the word that stands in the datasheets.
The nature of the scenarios is crucial to get at least an idea of \u200b\u200bwhat is actually going SSDs, they are "normal" or RAID configuration 0. It 's the reason why we chose to take the minimum and maximum among those registered, providing reference values \u200b\u200bcertainly not absolute (as related to our test suite), but related, to compare the units of analysis. We recreated the worst case scenario, or at least one of the most used, 100% writes, 100% random con pacchetti da 4K, prendendo in considerazione queue da 1, 8 e 32.
Quello che emerge è interessante. Si capisce subito che l'unico SSD 2,5 pollici della comparativa non nasce per compiti gravosi "da server ", poiché offre il suo massimo in questo scenario già con una queue singola. Completamente differente la situazione per quanto riguarda i modelli OCZ Revodrive, che fanno veramente la differenza in questo scenario. Si noti come il valore massimo registrato da OCZ Revodrive X2, a 4K 32 queue, sia quasi 5 volte superiore a quello del disco singolo e il 60% superiore rispetto a quanto registrato da OCZ Revodrive.
Non stupisca che il valore ottenuto da OCZ Revodrive X2 sia so far from the stated 18,434 against 120,000. The figure stated is always preceded by "up to", or "until". The value observed in this test on a OCZ Revodrive X2 is really high. To see the numbers go up we need to change test, to understand that when "until" is misleading or not.
So here we are to analyze the behavior in a very different scenario. Flowing the immense. Csv products from Iometer tests, we found the maximum values \u200b\u200bat the sequential read test 4K queue 16, which we propose a graph the results. OCZ
Revodrive showed a value close to 80,000 IOPS that, although still far declared by the 120,000, appear to be a major achievement. Proportionately more "honest" OCZ Revodrive model, with nearly 60,000 IOPS which is closer to 75,000 IOPS declared. Far, however, the MX-DS, which stops at around 10,000 IOPS, far from the 30,000 reported.
Considerations | | SSD Tests conducted on the comparative behavior showed a very different course depending on the model under consideration, recognizing, however, a measure of proportionality in most scenarios. The decision to insert a 2.5-inch disk with the same controller mounted on-board by Revodrive OCZ has allowed us to observe that the measure could affect Silicon Image RAID controller integrated models OCZ Revodrive, concluding not only that this is not a problem, but puts us in a lot of his handle extremely heavy loads for the the storage subsystem.
Easy slow fall into the trap of considering the model of Mach Xtreme Technology, especially if you come to conclusions with a glance to approximate graphs: tested definitely "abnormal" for a consumer SSD, the unit behaved very well, recording field values \u200b\u200bthat align with the best production 3Gbps SATA interface.
It 's the reason why, in these conclusions, we focus on the differences between the models and OCZ Revodrive Revodrive X2. The fans know that it is not just a symbol or the X2 to double stack memory would imply a doubling of performance. Just as in the case of SLI or Crossfire video cards in the world, a doubling of the physical components is by no means a guarantee of a doubling of the promised benefits. You can get closer, but knowing how much and how far you can climb performance turns out to be important if you are facing the idea of \u200b\u200bpurchase.
Well, what emerged from our test OCZ, with Revodrive X2, has done an excellent job. The difference with the model is evident Revodrive virtually every test, except in scenarios that limit the performance due to lack of load. Recall that OCZ Revodrive is equipped with two memory stacks, each controller SandForce SF-1200, offered in RAID 0 controller from Silicon Image, mind components Revodrive X2 Dual OCZ with the exception of the Silicon Image controller that remains unique. OCZ
Revodrive it is not the poor choice, since it allows to achieve performance very bright, much higher than those of a high-end SSD with SATA 3Gbps interface. All considerations are still empty and not relativized if we talk about prices. According to an online search on some shop have collected in a table of average prices as indicative, Free shipping and a large margin.
| | price | | Mach Xtreme MX-Technology 120GB DS | 190.00 to 200.00 € | | Revodrive 120GB OCZ | 290.00 to 310.00 € | | Revodrive X2 240GB OCZ | ~ € 550.00 | Several doubts that may arise getting ready to purchase a good SSD, if will be put in the crosshairs one of the units of analysis. If you think some time in an SSD controller with SandForce SF-1200 for your PC desktop is easy to fall into the temptation of OCZ Revodrive that, in the same capacity, it offers some extra € 100.00 (each decade discourse for obvious reasons if you think about an upgrade for your laptop ). € 100.00, but all in all, a significant difference be faced, especially in light of the results. If you already thought
directly to a PCIe SSD interface, doubts are still around the corner. While approximately € 300.00 you can bring home a Revodrive OCZ 120GB, you can not fail to notice that OCZ Revodrive X2, whereas the capacity double cost proportionately less, with a surplus of benefits that can make the difference in scenarios such as video editing, for instance. Proportionally, we said, because they serve more than € 500.00 anyway.
In any case, we have dealt with the devices that are intended to appeal to fans, which is why we believe that the reader is able to make their thoughts alone. In this article we wanted to offer some first-hand information on performance, with particular emphasis on the mutual differences among the models tested. |
| source: hardware upgrade | |
0 comments:
Post a Comment