Pentium G620 vs Liano...

[Kano]

@aceman

if your g620 does not support vt-x then it is disabled in the bios. see

http://ark.intel.com/products/53480

compared to i3 you dont have got ht and no quick sync (hardware h264 encoder). i3 vs. i5 means no aes commands thats similar to standard phenom vs. fx cpus, those fx cpus have got aes instructions. i5/i7 feature also a turbo mode which simpler cpus lack. usually i5 is not only dual (but there are some dual) but quad. i7 is usually quad + ht. the mobile variants are however dual without a Q in the name. somehow a bit confusing if you never looked at intel specs. for amd you just need to choose fx series when you want aes.

[aceman]

@aceman
if your g620 does not support vt-x then it is disabled in the bios. see

http://ark.intel.com/products/53480

Thanks, so it has VT-x and not VT-d. Must have misread the spec slightly.

compared to i3 you dont have got ht and no quick sync (hardware h264 encoder). i3 vs. i5 means no aes commands thats similar to standard phenom vs. fx cpus, those fx cpus have got aes instructions. i5/i7 feature also a turbo mode which simpler cpus lack. usually i5 is not only dual (but there are some dual) but quad. i7 is usually quad + ht. the mobile variants are however dual without a Q in the name. somehow a bit confusing if you never looked at intel specs. for amd you just need to choose fx series when you want aes.

I have no problem, I didn't want all those features and wanted silence instead (which I got). I just wanted to point out to the requester that it may not have many features. If he wants them he must look at the CPUs you pointed out.

[crazycheese]

But I really needed a first hand confirmation that it works, like "I got this MB and these DIMM and it rocks!"

I know that it should work in theory but as we all know, the difference between practice and theory is bigger in practice than in theory

My specific question are:

1. Does the MB need to support ECC, i.e. by providing the wiring to the DIMM slots? If we assume the processor supports ECC, it will need to be able to talk to the extra 8bits per 64-bit memory lane. I wonder if MB manufacturers provide the full wiring or they go cheap ass.
2. Provided that (1) is sorted out, can I use registered DIMMs in a non Opteron system? Or I need non buffered DIMMs? Can I mix those? (here I assume that buffered is synonymous with registered, please correct me if I'm wrong)

Cheers

http://www.google.de/search?q=amd+phenom+ecc

UPD: One good post: http://www.linuxquestions.org/questi...1/#post4066699

There are several confirmations that Asus boards work.

Also, I don´t understand why MB has to support ECC?.. Memory controller is in CPU, whats left are physicall lanes and DIMM sockets.
Buffered modules have buffer that ease the load on the controller and hence can allow you to install more modules. I have no idea, if AMD MB supports them, unbuffered ecc should work.

Maybe soon i will upgrade to ECC ram and answer that for you.

Also, I have athlon ii x4 630. Four cores at 2.8 Ghz.
It has WRONG voltage of 1.4v from factory - since it is PHENOM voltage and ATHLON does not have L3 cache.
Because AMD wanted to conceal some of PHENOMs as ATHLONs by simply disabling the L3, they left 1.4v as default, which is ultra stupid.
Athlon II x4 runs at 1.2v perfectly stable.

At 1.2v and with Scythe big schuriken, it gets 40C hot and unhearable in operation.

[kobblestown]

1. Yes the motherboard needs to explicitly support ECC RAM. AFAIK all Athlon and Phenom IIs support ECC, as should the FX series (Bulldozer). Bobcat (e.g. E-350) doesn't. Not sure about Llano. The system wont POST if the board doesn't support ECC but the CPU does and there is ECC RAM installed (I've had experience with this before).

Thanks for the info! Surprisingly, Asus don't advertise the ECC capability much. For instance, for the Asus M5A88-M the specification section on their Web site only says Non-ECC, whereas the manual says ECC/Non-ECC. I'd rather trust the manual because it also lists a BIOS option to control the ECC behaviour. In both cases it says Unbuffered so I guess I cannot use registered/buffered.

Apologies to the original poster for hijacking his thread. I hope he finds the ECC discussion useful since it clearly shows AMD has the advantage there, in case he is into ECC at all. I, for one, will always build my servers with ECC in the future + use Btrfs in mirror configuration. The data is all that matters. Please spare me the comments about how Btrfs is not ready for production. By the time I have enough money for a new server it will be

[devius]

I may be able to help you in the decision I hope. I have two systems that have some resemblance to what you're looking for.

One of those system is equipped with Athlon II X4 630, which basically has the same performance as the cpu part of A8-3800, installed on a 785G motherboard with 8GB ram and I'm using the HD4200 IGP. The other system is built around a Pentium G840 on a H61 motherboard and 4GB ram. Both of them are running Ubuntu 11.10. The Pentium is much more pleasant to work with mainly due to the better desktop experience provided by the GPU. For some reason (probably driver related) the composited desktop experience on the AMD system is terrible although the GPU is more than capable of handling it (I have to use Unity 2D to have a usable system). This is with both the open source and catalyst drivers. On proper fullscreen 3D applications the performance is as expected.

The Pentium is a totally different thing, with the desktop responding very smoothly and with very low latency. In terms of CPU power I don't really notice a difference because most applications are not well threaded enough. Even if they were it would only matter for certain workloads like 3D rendering. The problem here is not the CPU at all, but the GPU (and associated drivers).

For the type of workload you're describing, the Pentium will probably be the best choice. On top of the better experience you'll also be using open source graphics drivers so that should please the open source fanatics. Or not...

PS: The GPU on Llano is much more powerful than the HD4200 of my system, so the experience might be a bit better.

[s91066]

CON: only 2 cores, no HT, no virtualization (according to specs). Probably no overclocking.

Well, according to Intel (http://ark.intel.com/products/53480), the G620 has VT-x...
I am interested in a full load noise "test".
Could you please tell me if my "ready made" you mean a known branded PC (HP, Dell etc?)

@devplus: Thanks, your response is really helpful. If you imagine that on my PC the main problem is the desktop experience, (since the VM is a completely different thing), I think you helped me the most, as well as aceman.

Thanks

[t.s.]

Well, don't go for branded PC. It's more expensive.

Go for AMD X4 640 or G620. X4 if you want to maximize the core work (like VM, rip DVD, transcode, render, etc). Else, G620.

Agree with Devius. the open source radeon still no match for intel open source driver, for now. If you've spare money and go X4 630 route, take an cheap nvidia VGA card and you'll good to go.

[lem79]

Any of you with Athlon IIs, is there much lag in p-state switching? (i.e. takes a noticeable amount of time for the cores to change clock speed in response to a computational load)

Intel CPUs since Core 2 Duo I believe have very minimal latency in changing p-states, as do AMD CPUs since the original Phenom. The old Athlons up to the 65nm X2s took about 200ms to clock up in response to a load, which was noticeable in general usage. That *may* explain the difference between an Athlon II X4 and a Sandy Bridge Pentium on the desktop..

[devius]

I am interested in a full load noise "test".

Well, the Sandy Bridge Pentiums are in reality 35W TDP parts, although for marketing reasons (so they can sell more expensive "low-power" CPUs), they are branded as 65W. There are several sites out there that investigated the actual power consumption of these parts under full load and the results were between 31-37W. This means they don't get very hot and you can probably even underclock them, lowering power consumption even further and consequently noise as well.

[aceman]

Well, according to Intel (http://ark.intel.com/products/53480), the G620 has VT-x...

Yes, I was already corrected on that part, sorry.

I am interested in a full load noise "test".
Could you please tell me if my "ready made" you mean a known branded PC (HP, Dell etc?)
Thanks

Yes, it is a Lenovo IdeaCentre H420. I will try the full load test shortly.