*HP 2510p Owners Lounge*

[sfsilicon]

Originally Posted by radpilot

This discussion is very interesting.
Where can we see this compatibility list?

From Photofast. Try emailing them.

[sfsilicon]

Originally Posted by nando4

Presumably you are referring to the Photofast G-Monster V3 1.8" ZIF SSD compatibility list. I've echoed the msg I got from Photofast here, with them listing compatibility only some of the Panasonic CF-XX systems they tested. Saying that, I cannot see the sunplus sata-to-pata bridge chip + Indilinx controller they are using internally presenting too many issues, but I guess you never know until you try.

Would help if they tested Intel ICH2-8M I/O chipset with the UDMA5/ATA100 PATA interface and give some benchmarks. Systems that are likely to host their SSD.

Thanks NANDO4, there was a lot to read so I missed that.

I'm assuming they understand the ZIF connector spec better than Runcore. The HPminiuser forum is full of compatibility issues with connectors with their 1.8" ZIF PRO series.

I'd be willing to test the drive on a 2510p for a discount. They should also test it on the HP mini, MacAir v1 and Fujitsu Lifebooks that use 1.8" ZIF.

I'm just praying for a 1.8 ZIF drive SSD with similar or better power consumption as the current 1.8 4200 drive that I have and that will fully saturate the ATA channel with 100MB/s read/write instead of losing 12% performance to the bridge. $200 pricing for 64MB would be great.

[sfsilicon]

Not sure if this has been already posted, but this is why I am staying away from SSDs with Jmicron JMF602/B controllers. That goes also for the dual controller configurations.

Not sure if you saw this, but the worst case scenario for a JMF612B (2nd rev of the 612) on random write is 2042ms and average is 532.2ms. Which means that when a JMF612B drive gets used and runs out of freshly formatted banks a random write can take half a second and worst case 2s. This is why someone using a JMF612B SSD can see stuttering when they browse the web or use any application that writes small files to the disk frequently. All the highspeed sequential read/writes don't count when something like this happens frequently. Now there are programs to minimize this and Win7 will help if the Trim command is supported by the drive, but this is worse performance than a 4200rpm drive.

Iometer 4KB Random Writes, IOqueue=3, 8GB sector space

JMF602B MLC Drive
IOs per second = 5.61
MB/s = 0.02 MB/s
Average Latency = 532.2 ms (average 0.5s wait per write)
Maximum Latency = 2042 ms (average 2s wait per write)

Dual JMF602B MLC Controller Drive
IOs per second = 8.18
MB/s = 0.03 MB/s
Average Latency = 366.1 ms
Maximum Latency = 1168.2 ms

Intel X25-M (best of class)
IOs per second = 5923
MB/s = 23.1 MB/s
Average Latency = 0.51 ms

OCZ Vertex (2.5" MLC SATA Indilinx)
IOs per second = 1656
MB/s = 6.47 MB/ss
Average Latency = 1.81 ms

(source Anandtech)

I'm not sure why Indilinx picked 64MB. Cost could be a factor especially since the Indilinx controller is fairly expensive 30-50% more expensive than the JMF602B (in the $15 range). I suspect it might have to do more with performance vs overhead (NAND cell block size for erases, number of cells in parallel, how much buffering before cache flush) and managing the cache coherency especially at powerdown. Just as a reference most current SATA HDDs have a maximum of 32MB cache and the Indilinx has 2x that. It will be interesting to see what 256MB cache will do for performance and more importantly MLC NAND write cycle endurance. There is so much the tech industry has had to learn about MLC flash and making SSDs from them that fiascos like the JMF602 were bound to happen.

If you have a chance check out this Anandtech article. It is a long read, but really gives you a good background on the current state of the SSD technology. http://www.anandtech.com/storage/sho...spx?i=3531&p=1

[pocket_geek]

So I though I'd post some comparison benchmarks for how I have my 2510p configured right now. It currently has my Samsung 64GB 1.8" ssd installed as master and also has a Western Digital 320GB pata drive installed as slave in the optical drive bay via the Newmodeus drive caddy and a pata circuit board.

On the left we have the Samsung and on the right we have the Western Digital:

[sfsilicon]

Originally Posted by pocket_geek

On the left we have the Samsung and on the right we have the Western Digital:

The cache (16MB or 32MB?) on the HDD makes quite a difference for the 4k R/W numbers. Thanks for sharing.

[pocket_geek]

8MB of cache on that drive...still makes a pretty big difference though.

[pocket_geek]

I would have rather used the sata circuit board and a 500GB drive, but it impacted my battery life too much.

[sfsilicon]

Originally Posted by nando4

Without no disclosure from HP on what that 24-pin socket in the 1.8" drive bay is for, the least obtrusive method of perhaps finding out if it is a SATA connector is by comparison against the pinout of a 2730P 24-pin SATA connector. The theory to do this is explained at the bottom of this post titled "SUMMARY: how to get the unknown connector working as a SATA connector, if it is one". Without bios support, bootup might be another issue as explained there too *if* that secondary IDE controller that appears in Device Manager is not the SATA device running in emulated mode.

So the 2730p pinout is not documented in anyway? If we had that it would be easier to guess which were the relevant pins then map it to the socket.

BTW, does anyone know what the socket type is in the 2510p/2710p? That would give us an idea where pin 1 is and in which direction the higher pins are oriented.

Originally Posted by nando4

So then without native SATA ability then the 2510P practical SSD options are: 1.8" ZIF SSD, 1.8" SATA SSD using DIY microsata-to-ZIF adapter (cramped) or 2.5" SATA SSD via sata-to-pata optical bay caddy.

I'm going to test the ebay pata-sata caddy. That has a reasonable chance of success and coupled with an Indilinx SSD maybe decent performance with higher power consumption. I messaged the seller on eBay a few days ago about 2510p support and have not gotten any response. Not a good sign in general. I'll share results when I have a chance.

BTW, I got my Intel 4965AGN mini PCIe wireless card with Draft N support. The card was easy to install and I just DL the drivers from Intel.com. The card works in g mode, but I need to add a third antenna to enable the N mode. Strangely the Intel drivers do not allow a manual override to N. Just a tip make sure you buy a card that has the extra antenna to avoid the hassle. Now I'm going to see if I can easily reroute one of the antennas for the wireless broadband to the WiFi port.

[sfsilicon]

Originally Posted by pocket_geek

I would have rather used the sata circuit board and a 500GB drive, but it impacted my battery life too much.

Hey Pocket_Geke what impact are you seeing on your battery life? What kind of battery do you have (3/6/9 cell?)

[nando4]

Originally Posted by sfsilicon

So the 2730p pinout is not documented in anyway? If we had that it would be easier to guess which were the relevant pins then map it to the socket.

BTW, does anyone know what the socket type is in the 2510p/2710p? That would give us an idea where pin 1 is and in which direction the higher pins are oriented.

2730 24-pin 1.8" SATA connector pinouts = unknown.
2530P 8-pin 1.8" SATA connector pinouts = unknown.

microsata only needs 6 wires to run, so likely only using a few of those pins in that 24-pin socket anyway. Only way of mapping it's pins is a close up view of the tracks OR to get the cable and do continuity probes against the ends. The 2510P socket itself has little numbers against the pin numbers which I've shown at the bottom of this post.

Originally Posted by sfsilicon

I'm going to test the ebay pata-sata caddy. That has a reasonable chance of success and coupled with an Indilinx SSD maybe decent performance with higher power consumption. I messaged the seller on eBay a few days ago about 2510p support and have not gotten any response. Not a good sign in general. I'll share results when I have a chance.

Look forward to hearing your results. As mentioned, I believe this thread and the 6510b optical bay caddy thread has had inaccurate reporting of the ebay sata-to-pata caddy's performance. A credible 2510P owner can perhaps give a more accurate report. Comparatively, the ebay caddy looks to be more of a flush-fit and I'm curious whether the Marvell sata-to-pata bridge might be slightly better performing as shown by the mtron-in-sata-to-pata ultrabay performance here. Ultrabay and ebay caddy both use the same Marvell bridge chip as shown here and here.