Windows upgrade on older hardware, SSD or not?

[mikeymikec]

The board is an ASUS M2N-SLI Deluxe, which has an nforce chipset (which can't do AHCI) and a JMicron RAID controller (which can do AHCI).

I'll probably go for Windows 7 64 on it (from XP) rather than Win81 as I think that'll be less of a compatibility risk.

The budget isn't that tight, the computer runs pretty well generally speaking. I can't decide whether it's worth risking trying an SSD in there or play it say and get a WD Black for the best HDD performance.

RAM will get upgraded to 4GB.

 

[Insert_Nickname]

The board is an ASUS M2N-SLI Deluxe, which has an nforce chipset (which can't do AHCI) and a JMicron RAID controller (which can do AHCI).

You sure about that? Far as I know, nForce SATA controllers work just fine with the standard MS AHCI driver included in Vista/7/8.

IDE mode will kill SSD performance...

I'll probably go for Windows 7 64 on it (from XP) rather than Win81 as I think that'll be less of a compatibility risk.

Since its an AM2(+) CPU, x64 8(.1) should work. The only Athlon64's to have issues are Socket 939 and 754 ones. You can double check quickly, if your CPU is an F2 revision or newer, there shouldn't be problems running 8(.1). Else stick to 7.

The budget isn't that tight, the computer runs pretty well generally speaking. I can't decide whether it's worth risking trying an SSD in there or play it say and get a WD Black for the best HDD performance.

RAM will get upgraded to 4GB.

See above. It'll depend on weather or not you can get AHCI working...

 

[mikeymikec]

AHCI is only mentioned in the manual (latest download from the ASUS site) with regard to the JMicron controller.

 

[mfenn]

AHCI is only mentioned in the manual (latest download from the ASUS site) with regard to the JMicron controller.

Have you updated the BIOS to the latest and look at the BIOS options?

 

[Leyawiin]

IDE mode will kill SSD performance...

570 chipset doesn't support AHCI - tried it recently with a Gigabyte AM2 570 board last spring (paired with an Athlon X2 6400+). It had a SSD in IDE mode and while it wasn't as fast as in AHCI mode it was still miles better than the old 80GB hard drive that was in it.

 

[VirtualLarry]

While it's true that older NV chipsets don't properly support AHCI, an SSD even in IDE mode will provide massive performance improvements over any mechanical HDD.

Just get an SSD, use your old system (now accelerated), and enjoy!

 

[Insert_Nickname]

570 chipset doesn't support AHCI - tried it recently with a Gigabyte AM2 570 board last spring (paired with an Athlon X2 6400+). It had a SSD in IDE mode and while it wasn't as fast as in AHCI mode it was still miles better than the old 80GB hard drive that was in it.

It doesn't? Thanks for the heads-up. That is good to know. Strangely enough I've had a Geforce 6150/430 (MCP51) board that did support ACHI. That's an even older chipset. Strange.

While it's true that older NV chipsets don't properly support AHCI, an SSD even in IDE mode will provide massive performance improvements over any mechanical HDD.

Just get an SSD, use your old system (now accelerated), and enjoy!

Oh, I should have quantified that statement. Yes, an SSD will be faster then any HDD, but there is absolutely no need to get a high-performance SSD if you're going to run it in IDE mode. Any SSD will do, even an X25-V 40GB...

 

[mikeymikec]

Ok, I went for it and I've been banging my head against this problem for a good hour or so.

I've updated to the latest release version of the BIOS, connected the SSD to the JMicron SATA port, enabled the controller in the BIOS, set it to AHCI mode, but there's no mention during POST of the controller, or any "JMicron BIOS" screen appearing before the computer attempts to boot.

I've tried setting the controller to RAID, IDE or AHCI, but the JMicron BIOS doesn't appear. I've tried using Ctrl+J (mentioned in the manual to access the JMicron BIOS) lots during POST in the hope that it would bring up its configuration, but nothing.

I've tried disabling the other storage controllers (IDE, SATA IDE), still nothing.

The SSD detects fine if I connect it to one of the nvidia SATA ports, but even if it didn't, I'd still expect some mention of the JMicron BIOS in order to boot from it. The BIOS's boot device listing doesn't mention it at all, when I've got the SSD connected to the JMicron port, the boot device listing acts like I don't have any hard drives connected.

I have a fallback option of a PCIE SATA AHCI controller, but I feel like I have to be doing something stupid for this not to be working. My only other thought is perhaps ASUS disabled the JMicron controller in some way (intentionally or not) through a BIOS update, however I had to update the BIOS in order to get 4GB RAM working.

I've tried resetting the BIOS twice. I haven't tried removing the battery, yet.

 

[mikeymikec]

I feel like breaking something now, I can't believe what's just happened.

I resigned myself to putting in the PCIE AHCI controller card I have, which I did, and I connected up the SSD and DVD drive to that controller. I switched on the computer, then I saw the JMicron BIOS appear, then the PCIE AHCI controller BIOS!

I then removed the card, set the BIOS back the way I wanted it, and the sodding JMicron BIOS is appearing as it should, with the SSD detected.

The only vague possibility I can think of to explain this turn of events is that the computer needs to be powered down after the JMicron controller is enabled, but I'm not sure I buy that because I think it was enabled before I added the 4GB RAM after the BIOS update.

 

[mikeymikec]

FYI for anyone thinking of using the JMB363 AHCI controller with an SSD:

1 - It isn't recognised by Win7 as a standard AHCI controller (which makes me worry about things like TRIM) despite being configured as AHCI in the BIOS.

2 - The performance is pretty pants, like SATA 1.5GBps when it should be in the 3Gbps zone. This is with the latest JMB36x driver from JMicron. Benchmarked with ATTO.

I switched over to the PCIE AHCI SATA 6Gbps controller made by Startech. Marvell controller. With the latest driver from the Startech website, there was a performance improvement compared to the MSAHCI driver, certainly in the SATA 3Gbps territory (read speeds peaked at about 220MB/sec). Better than the JMicron controller all round, but it's a bit of a lie to claim it is 6Gbps capable. This is with a Samsung 840 PRO 128GB SSD, which I know exactly what to expect as I've used it with Ivy Bridge and Haswell chipsets (they tear strips off the Marvell controller, easily making good use of the 6Gbps bandwidth).

 

[mikeymikec]

Ah, I just found this on the Startech site:

The maximum throughput of this card is limited by the bus interface. If used with PCI Express Gen 1.0 enabled computers, the max throughput is 2.5 Gbps. If used with PCI Express Gen 2.0 enabled computers, the max throughput is 5 Gbps. Only one port can use the Port Multiplier feature at a time.

 

[Insert_Nickname]

The only vague possibility I can think of to explain this turn of events is that the computer needs to be powered down after the JMicron controller is enabled, but I'm not sure I buy that because I think it was enabled before I added the 4GB RAM after the BIOS update.

I'm not familiar with that BIOS implementation, but isn't there a setting to force the BIOS to show OROMs during POST somewhere? By plugging in the PCIe card, you may have forced that setting to enable/on...?

Ah, I just found this on the Startech site:

The maximum throughput of this card is limited by the bus interface. If used with PCI Express Gen 1.0 enabled computers, the max throughput is 2.5 Gbps. If used with PCI Express Gen 2.0 enabled computers, the max throughput is 5 Gbps. Only one port can use the Port Multiplier feature at a time.

That's right. 220MB/s sound about right for a PCIe 1.1 x1 controller. If you use it in a 2.0 slot, expect 350-400MB/s on a good day.

 

[mikeymikec]

I looked for a setting to load the JMicron ROM but didn't find anything.

 

[mfenn]

I
That's right. 220MB/s sound about right for a PCIe 1.1 x1 controller. If you use it in a 2.0 slot, expect 350-400MB/s on a good day.

The JMicron internal controller is likely connected via PCIe 1.1 internally as well, so it's probably of no advantage in terms of card connectivity.

 

[mikeymikec]

It was slower than the controller card, possibly almost to the point where a WD Black could have competed with it based purely on ATTO results (120MB/sec peak IIRC).

My guess is that PCIE bandwidth and substandard drivers were limiting it. I just wonder how well it would have done in the capacity it was supposedly originally for, ie. RAID 0 or 1 with hard disks from that time. I wonder if it was a gimmick, as I'm pretty sure that the NVIDIA SATA 3Gbps controller in RAID mode would have torn strips off it (at least, based on my experience of that controller with a single WDB drive).

 

[mikeymikec]

Meh... the Marvell driver can't handle sleep mode (0x0000007E SYSTEM_THREAD_EXCEPTION_NOT_HANDLED during resume). It is fine with hibernation apparently.

My conclusion from this whole experience is fairly similar before I considered this whole escapade: If the native SATA implementation can't do AHCI, then don't bother with an SSD. Unless you dislike playing it safe

I'm going to check the BIOS for any odd options with regard to sleep mode, but I doubt I'll find anything useful.

- edit - Nothing interesting found. I'm going to try S1 sleep mode so that if the machine was inadvertently put into sleep mode, hopefully it won't crash on resume.

 

[VirtualLarry]

If the native SATA implementation can't do AHCI, then don't bother with an SSD.

And you've benchmarked daily tasks, with the SSD installed as IDE, versus AHCI, and determined that the SSD doesn't perform well in IDE mode? Because I've got one rig installed in IDE mode, and I haven't noticed much slowdown. Most desktop workloads, other than booting, are QD=1.