|
am I reading this wrong? how are my numbers so far off what i see in the reviews? I know there is the thing going on where components are being replaced without telling people but this seems way off compared to the crystaldiskmark numbers here https://www.tweaktown.com/reviews/9447/adata-xpg-sx8200-pro-2tb-2-ssd/index.html.
|
#
?
Dec 10, 2020 18:05
|
|
|
|
| # ? May 13, 2024 12:44 |
|
|
You must have some ram cache on, pcie3 x4 maxes out at 3.5GB/s
|
|
#
?
Dec 10, 2020 18:12
|
|
|
Perplx posted:You must have some ram cache on, pcie3 x4 maxes out at 3.5GB/s yeah thats it, thanks. ill retest here we are  Just based on the #s would you say this ssd is similar to the review copies in spec, or is it one of the negatively affected ones? It was purchased a couple months ago Fauxtool fucked around with this message at 18:29 on Dec 10, 2020 |
|
#
?
Dec 10, 2020 18:19
|
|
|
Close, those RND4K are way way over the top. (what hell you got goin on there?!)
|
|
#
?
Dec 10, 2020 20:45
|
|
|
Im on an asus b-550e and 3800xt. i turned off ramcache. I've no idea what else could be causing that.
|
|
#
?
Dec 10, 2020 21:11
|
|
|
Fauxtool posted:Im on an asus b-550e and 3800xt. Whats 'ramcache'?
|
|
#
?
Dec 10, 2020 21:37
|
|
|
its a utility on some asus MOBOs that lets you use extra ram as a scratch disk sort of like intel optane.
Fauxtool fucked around with this message at 21:54 on Dec 10, 2020 |
|
#
?
Dec 10, 2020 21:49
|
|
|
Fauxtool posted:its a utility on some asus MOBOs that lets you use extra ram as a scratch disk sort of like intel optane. Oh! Makes sense.
|
|
#
?
Dec 10, 2020 23:01
|
|
|
Fauxtool posted:yeah thats it, thanks. ill retest Your random 4k test is using a different queue depth than the test they ran in the review you linked, so the results are not directly comparable. Queue depth changes results a lot. The change in controller and nand is really not worth being super concerned about for average desktop apps and games users. Basically it's enough of a difference to be annoyed at Adata for stealth changes to components, but not enough to make it a poor performer. You would never be able to tell in normal use without benchmarks. (If you are using it for high-performance work, the major drop in sustained 4k IO is bad. This is IMO a reason for people to start avoiding Adata, especially professional users, because it's a bush league move.)
|
|
#
?
Dec 10, 2020 23:38
|
|
|
thats a relief, thanks
|
|
#
?
Dec 10, 2020 23:57
|
|
|
I mean if thats real IOPs, thats some chad poo poo right there. badass drive!.
|
|
#
?
Dec 11, 2020 00:50
|
|
|
There's nothing too magical about those IOPs given the 4K random.. Just divide 851MB/s by 4096 and you get 207k. IOPS = IOs (at 4k) per second given the 851MB/s It's good tho!
|
|
#
?
Dec 11, 2020 03:04
|
|
|
Why does crystal disk mark have an underage looking anime girl
|
|
#
?
Dec 11, 2020 05:51
|
|
|
The "Shizuku" edition does, but guessing from Fauxtool's avatar it doesn't bother him
|
|
#
?
Dec 11, 2020 05:54
|
|
|
TITTIEKISSER69 posted:. Its a japanese dev and the anime versions are the "real versions." That's right versions. There is also the kurei kei one and each has many many skins built in Shes the company mascot and has been for almost a decade Fauxtool fucked around with this message at 06:10 on Dec 11, 2020 |
|
#
?
Dec 11, 2020 06:06
|
|
|
redeyes posted:I mean if thats real IOPs, thats some chad poo poo right there. badass drive!. priznat posted:There's nothing too magical about those IOPs given the 4K random.. And that's why I said that queue depth is important. Increasing queue depth allows the controller to pick which order it handles requests; at high queue depth the controller has lots of room to find the most optimal order for performance. OTOH you also get increased latency per command, since read command #1 might not be responded to until much later. Unfortunately there aren't many real-world things modeled by 4k random / QD32. It's the type of test a drive maker will claim IOPS for their paper spec. Real-world use of highly random reads tends to have low queue depth (though possibly multiple threads). So if you're looking at IOPS as measured by review websites. which tend to use either low queue depth or are doing real-world performance capture for their benchmarks, no that 200k result isn't "real". (And none of this matters for home users since the average desktop use pattern in no way resembles a 4k random test. If you need a fast drive for video games, you care about sequential read and not much else.)
|
|
#
?
Dec 11, 2020 06:45
|
|
|
All this means is that you should buy Optane and lots of it
|
|
#
?
Dec 11, 2020 06:51
|
|
|
My motherboard's m.2 slot pretty much gets at least half covered up by my dual-slot-width gpu. Would it be a bad idea to use an m.2 drive in this scenario, heat-wise?
|
|
#
?
Dec 11, 2020 21:54
|
|
|
Also when doing random write tests on NAND SSDs you ideally want to fill up the drive beforehand, then start doing random writes for some time and THEN actually check the performance. The results with/without garbage collection in the drive can be pretty big as your writes are going to be contending with internal read/write/erases, or just going straight to the media when the drive is mostly empty, or not really internally 'defragmented'. Not that it really should matter for your average use case, you don't do enough 4k random writes to really notice the shift in performance.
|
|
#
?
Dec 11, 2020 21:57
|
|
|
taiyoko posted:My motherboard's m.2 slot pretty much gets at least half covered up by my dual-slot-width gpu. Would it be a bad idea to use an m.2 drive in this scenario, heat-wise? SSDs don't generate much heat and are perfectly happy running a little warm. Coming at it from another perspective, your airflow isn't worse then the standard laptop M.2 slot and nobody has ever complained about their SSD overheating there.
|
|
#
?
Dec 11, 2020 23:15
|
|
|
Yeah, NAND is actually quite happy when it's hot - the only thing that could use some extra cooling on the drive itself is the controller chip. The DRAM cache too, maybe. The only time I've ever been  about any kind of memory is when I ran 4x4GB on my last build. The stick in DIMM0 always went bad after a few months' time and I figured it had to be the fact that it was in a heat trap under my big Silver Arrow HSF. Switched to 2x8 in DIMM1 and 3 and never had a problem from then on.
|
|
#
?
Dec 12, 2020 01:44
|
|
|
New Optane around the corner https://www.zdnet.com/google-amp/article/intel-claims-fastest-data-centre-storage-title-with-optane-charged-p5800x/ quote:According to the spec sheet, the P5800X is capable of hitting 7.2GB per second on sequential reads, 6.2GB per second on sequential writes, up to 1.5 million IOPS of random 4k reads and writes, which can increase to 1.8 million IOPS for mixed workloads of 70% reads and 30% writes. https://www.anandtech.com/show/16318/intel-announces-new-wave-of-optane-and-3d-nand-ssds quote:Intel has also made optimizations for single-sector 512-byte random reads, which can hit 4.6M IOPS WhyteRyce fucked around with this message at 17:10 on Dec 16, 2020 |
|
#
?
Dec 16, 2020 16:42
|
|
|
hey WTF are these: https://www.intel.com/content/www/us/en/products/memory-storage/optane-memory/optane-memory-h10-solid-state-storage.html Anyone know anything?
|
|
#
?
Dec 16, 2020 17:41
|
|
|
redeyes posted:hey WTF are these: https://www.intel.com/content/www/us/en/products/memory-storage/optane-memory/optane-memory-h10-solid-state-storage.html Desperation to unload optane 16/32GB ICs most likely 
|
|
#
?
Dec 16, 2020 19:08
|
|
|
As far as I can tell, they're just a 16/32GB optane module and QLC SSD on the same m.2 stick. Presumably the optane is used in place of a more traditional DRAM or SLC cache. https://www.pcworld.com/article/3389742/intel-optane-memory-h10-ssd-review.html
|
|
#
?
Dec 16, 2020 19:13
|
|
|
It's that Optane based software caching Intel pushed out awhile ago but now on a single M.2 device instead of requiring two separate drives. Basically throw a QLC nvme drive and an Optane nvme drive together and have the Intel RST software handle managing it Putting it on a single device makes it a slightly better value proposition to OEMs. I don't know if it's worth it for most people unless the real world performance is significantly better than just buying an equivalent size TLC drive or if the price is competitive with equivalent QLC sized drives. Or both. WhyteRyce fucked around with this message at 20:21 on Dec 16, 2020 |
|
#
?
Dec 16, 2020 20:18
|
|
|
Does it actually enumerate as two NVME devices which have to be managed by Intel RST, or is it a single NVME device whose own firmware manages where blocks are stored?
|
|
#
?
Dec 16, 2020 23:38
|
|
|
BobHoward posted:Does it actually enumerate as two NVME devices which have to be managed by Intel RST, or is it a single NVME device whose own firmware manages where blocks are stored? I believe it should appear as two separate NVMe devices if you don't have the RST muckery installed or turned on.
|
|
#
?
Dec 17, 2020 00:35
|
|
|
Yep that PC World article says they show up as two drives, each controller is wired to two lanes of the m.2 slot TIL motherboards just let you bifurcate m.2 slots like that
|
|
#
?
Dec 17, 2020 00:41
|
|
|
repiv posted:Yep that PC World article says they show up as two drives, each controller is wired to two lanes of the m.2 slot If that's the case, while the Optane-backed portion might have real good latency performance, bulk transfer rates are gonna be considerably lower than normal SSDs. Might make sense if you've got a database or something that you can get real high hit % with within the Optane size, but otherwise it makes it a pretty weird device.
|
|
#
?
Dec 17, 2020 00:44
|
|
|
AT reviewed it when they came out last year: https://www.anandtech.com/show/14249/the-intel-optane-memory-h10-review-two-ssds-in-one The last sentence: quote:The combination of QLC and Optane might still be able to produce an impressive consumer product, but it will take more work from Intel than this relatively low-effort product. Ouch.
|
|
#
?
Dec 17, 2020 00:45
|
|
|
Wow I had figured it was going to be like a dram replacement on the ssd, that’s very odd it is its own x2 device. Gotta be for OEMs/Laptops only because this would be a recipe for disaster if people are just buying it off newegg then wondering why they only get 16GB storage on their m.2 slot that doesn’t bifurcate.
|
|
#
?
Dec 17, 2020 00:47
|
|
|
BobHoward posted:Does it actually enumerate as two NVME devices which have to be managed by Intel RST, or is it a single NVME device whose own firmware manages where blocks are stored? It's two devices sharing one PCIe bus and has to be managed by a PCIe controller that can bifrucate x4 to 2x2. It's usable only with recent Intel southbridge chips, which they seem to have designed with this type of job in mind. The RST software is required to use it as a unified drive. And the software caching doesn't do as good a job melding Optane and QLC as conventional SLC-TLC drives. It's not a good drive and was probably only appealing to OEMs who were getting it at discount (and using the optane to fluff up their system specs by adding it to system memory).
|
|
#
?
Dec 17, 2020 00:50
|
|
|
Needing some Intel software running seems like a sure fire way to have it blow up in your face and lose data
|
|
#
?
Dec 17, 2020 00:52
|
|
|
DrDork posted:If that's the case, while the Optane-backed portion might have real good latency performance, bulk transfer rates are gonna be considerably lower than normal SSDs. Might make sense if you've got a database or something that you can get real high hit % with within the Optane size, but otherwise it makes it a pretty weird device. Continuous bulk transfer isn't really good use case with this because you'll probably blow through the Optane cache pretty easily and the data isn't going to stay there. If I remember the first batches of reviews for Optane Accelerators, the initial first time load for games was slow NAND speeds and once there then you got the acceleration benefits. priznat posted:Wow I had figured it was going to be like a dram replacement on the ssd, that’s very odd it is its own x2 device. Gotta be for OEMs/Laptops only because this would be a recipe for disaster if people are just buying it off newegg then wondering why they only get 16GB storage on their m.2 slot that doesn’t bifurcate. Cygni posted:Ouch. Yeah they just grafted together some stuff to quickly get something out the door from the look and feel of it. Designing a brand new type of controller to work with a giant Optane cache in harmony with a giant block of NAND probably isn't a quick job and probably not something they want to dump a ton of resources in for a razor thin margin market. WhyteRyce fucked around with this message at 01:01 on Dec 17, 2020 |
|
#
?
Dec 17, 2020 00:53
|
|
|
Thanks Ants posted:Needing some Intel software running seems like a sure fire way to have it blow up in your face and lose data I bought a 16GB drive for $9 off Ebay to mess around with. You have to enable something in BIOS to get some hooks in the driver to load (setting is part of some overloaded option from what I remember on my board, although that probably depends on the vendor) but I couldn't get it to detect with the RST software. After blindly and continuously messing around with the settings I think I might have got the first stage to kick in but then I wasn't able to boot Windows. I'm assuming it goes smoother if it's some OEM system where everything is pre-configured, or if you are doing it yourself from scratch and don't have to worry about blowing away your Windows install. WhyteRyce fucked around with this message at 01:01 on Dec 17, 2020 |
|
#
?
Dec 17, 2020 00:57
|
|
|
Man that is just the laziest way of doing it. In order of decreasing  , here's their options IMO- single x4 NVMe controller, controller manages data placement - single x4 NVMe controller, controller provides two NVMe namespaces, one for the QLC and one for the Optane. (NVMe namespaces are somewhat similar to SCSI Logical Unit Numbers, if anyone remembers that poo poo.) - just throw two preexisting SSD designs onto a single PCB layout through bifurcation and call it a day Ironically, from a technological point of view, the match made in heaven for this device is Apple's Fusion Drive. It was originally designed to accelerate off the shelf standard HDDs with a much smaller off the shelf standard SSD, and would do fine doing the same if the "HDD" was a slow QLC SSD and the "SSD" was a fast Optane SSD. Too bad for Intel that Apple's been designing their own SSD controllers for years and probably has negative interest in Optane until it's a real mass market product.
|
|
#
?
Dec 17, 2020 08:08
|
|
|
A single controller to control both is actually really difficult. I think Intel doesn't even have a consumer level in-house built NAND controller they could tweak. The Optane and NAND controller are different enough that you can't just glob on some extra logic to do both and the FW would probably be a nightmare. Optane was bleeding NSG enough I doubt they wanted to spin up another ASIC and FW team and all that a new product entails. A single controller is the more compelling product and maybe what they should have went in on first in the consumer market but hey it's a new tech trying to build a new market so I can empathize with taking a couple of smaller bets. They still are on first generation controllers after all and I don't think they wanted to bite off more than they could chew. WhyteRyce fucked around with this message at 08:38 on Dec 17, 2020 |
|
#
?
Dec 17, 2020 08:20
|
|
|
WhyteRyce posted:They still are on first generation controllers after all Exclusive look at the next generation: 
|
|
#
?
Dec 17, 2020 16:47
|
|
|
|
| # ? May 13, 2024 12:44 |
|
|
I don't even know if there will be a hybid Optane drive in the future. I always thought the client business at Intel NSG was primarily to find some kind of use for all those NAND wafers their factories were spitting out that couldn't be sold in a much more lucrative enterprise product. Once the acquisition completes they'll have to go out and buy someone else's NAND in order to throw together a product and at that point why bother.
|
|
#
?
Dec 17, 2020 19:15
|
|