TR Forums

OK the G4620 and i3-7300 CPU seems awful  close together in features and clockspeed, so why buy one over the other?

The i3-7300 has 1MB more (L2/L3??) cache and supports VT-d. If you're running VMs and want to pass native I/O through the hypervisor, take the 7300. Not sure the exact effect of the extra cache, but more cache is always better. For $50, I'd take the 7300 on general principles.

Cache and 300 MHz of clock speed...

Captain Ned wrote:

The i3-7300 has 1MB more (L2/L3??) cache and supports VT-d.  If you're running VMs and want to pass native I/O through the hypervisor, take the 7300.  Not sure the exact effect of the extra cache, but more cache is always better.  For $50, I'd take the 7300 on general principles.

For VT-d alone, I would get the i3-7300 for myself for a cheap gaming box. But what about for general non-tech user?

whm1974 wrote:

For VT-d alone, I would get the i3-7300 for myself for a cheap gaming box. But what about for general non-tech user?

Take the cache and the clock speed. That alone is about another year of "future-proofing". After all, we're talking about at most a $50 price delta. The G4620 will probably OC better, but that's not anything that a "beige box" user needs.

Captain Ned wrote:

whm1974 wrote:

For VT-d alone, I would get the i3-7300 for myself for a cheap gaming box. But what about for general non-tech user?

Take the cache and the clock speed.  That alone is about another year of "future-proofing".  After all, we're talking about at most a $50 price delta.  The G4620 will probably OC better, but that's not anything that a "beige box" user needs.

Yeah that makes sense.

whm1974 wrote:

For VT-d alone, I would get the i3-7300 for myself for a cheap gaming box. But what about for general non-tech user?

That doesn't make any sense. I don't think a "cheap gaming box" is going to include running Windows in a VM, and if it does, it shouldn't. That just doesn't make any sense.

DancinJack wrote:

whm1974 wrote:

For VT-d alone, I would get the i3-7300 for myself for a cheap gaming box. But what about for general non-tech user?

That doesn't make any sense.  I don't think a "cheap gaming box" is going to include running Windows in a VM, and if it does, it shouldn't.  That just doesn't make any sense.

Not for someone who going to use VMs awful lot. But for a user who just want to check out Linux distros from time to time, well maybe?

whm1974 wrote:

Not for someone who going to use VMs awful lot. But for a user who just want to check out Linux distros from time to time, well maybe?

Why would you need VT-d for "checking out Linux distros from time to time?"

DancinJack wrote:

Why would you need VT-d for "checking out Linux distros from time to time?"

whm1974 might, as that's his thing. Joe Beige Box doesn't know and doesn't care, but the extra cache and clock is something he will directly notice.

Captain Ned wrote:

DancinJack wrote:

Why would you need VT-d for "checking out Linux distros from time to time?"

whm1974 might, as that's his thing. Joe Beige Box doesn't know and doesn't care, but the extra cache and clock is something he will directly notice.

He wouldn't though! You don't need VT-d for "trying out Linux distros" from time to time and chances are you would never even notice the CPU has it if that's what you were doing.

I have no disagreement on the clock speed and cache. All for those.

Captain Ned wrote:

DancinJack wrote:

Why would you need VT-d for "checking out Linux distros from time to time?"

whm1974 might, as that's his thing.  Joe Beige Box doesn't know and doesn't care, but the extra cache and clock is something he will directly notice.

Good point. Every so often I get folks asking me for computer advice. Basically, I just give them some decent specs to look for.

DancinJack wrote:

He wouldn't though! You don't need VT-d for "trying out Linux distros" from time to time and chances are you would never even notice the CPU has it if that's what you were doing.

Absolutely correct. VT-x is the "virtualization extensions" that pass the CPU(S) on to the host's guests. This is pervasive for nearly a decade. VT-d is "directed I/O" for passing PCIe devices (and PCIe devices only!) on to a host's guests. This is the one missing from 4770K and earlier top-end models (amongst others) for the infamous product segmentation.

I'm also in agreement with the clock speed and cache for any serious work. Although just last night I was mulling i3 vs Pentium for BlueIris. Broadwell NUC i3-5010U wasn't powerful enough; desktop Skylake i3-6100 with same workload is nearly idle (~6%).

AVX 2.0 on i3
no AVX at all on pentium

VT-d != VT-x

For a cheap gaming machine both VT-d and AVX 2.0 make no difference. The more cache might down the line, but I find the $50 difference too much in this category. Saying that, the G4620 is not the one to look at here, the G4560 for $64 would be a better choice. That's half price of the i3-7300 and the money saved covers the 8GB RAM and some of the H110 motherboard costs. Pair that with a 1050 or a 1050Ti and you have a nice entry level gaming PC.

Edit: you only lose 200Mhz compared to the G4620 (3.5Ghz vs. 3.7Ghz), so 5-ish %

whm1974 wrote:

DancinJack wrote:

whm1974 wrote:

For VT-d alone, I would get the i3-7300 for myself for a cheap gaming box. But what about for general non-tech user?

That doesn't make any sense.  I don't think a "cheap gaming box" is going to include running Windows in a VM, and if it does, it shouldn't.  That just doesn't make any sense.

Not for someone who going to use VMs awful lot. But for a user who just want to check out Linux distros from time to time, well maybe?

To be fair, he said "for myself."   Because VM usage is a .0001% activity for a home PC.

Kretschmer wrote:

whm1974 wrote:

DancinJack wrote:

That doesn't make any sense.  I don't think a "cheap gaming box" is going to include running Windows in a VM, and if it does, it shouldn't.  That just doesn't make any sense.

Not for someone who going to use VMs awful lot. But for a user who just want to check out Linux distros from time to time, well maybe?

To be fair, he said "for myself."   Because VM usage is a .0001% activity for a home PC.

Well I do other things with my system besides playing games as well. Speaking of which, I should get off my rear and learn how to use Gimp, LibreOffice, and what not.

whm1974 wrote:

Well I do other things with my system besides playing games as well. Speaking of which, I should get off my rear and learn how to use Gimp, LibreOffice, and what not.

Learn how to use? If you haven't used them by now, I don't know that you have a need.

The Egg wrote:

whm1974 wrote:

Well I do other things with my system besides playing games as well. Speaking of which, I should get off my rear and learn how to use Gimp, LibreOffice, and what not.

Learn how to use? If you haven't used them by now, I don't know that you have a need.

Does seem rather odd for someone who has been so vocal about Linux and Open Source...

just brew it! wrote:

The Egg wrote:

whm1974 wrote:

Well I do other things with my system besides playing games as well. Speaking of which, I should get off my rear and learn how to use Gimp, LibreOffice, and what not.

Learn how to use?  If you haven't used them by now, I don't know that you have a need.

Does seem rather odd for someone who has been so vocal about Linux and Open Source...

I sort of played a around with them, but I haven't done anything major with them.

I am willing to spend more for a Core i3 to get the AES-NI support, VT-x support, greater cache, etc.

Aether wrote:

I am willing to spend more for a Core i3 to get the AES-NI support, VT-x support, greater cache, etc.

The G4620 has VT-x and AES-NI.

Aether wrote:

I am willing to spend more for a Core i3 to get the AES-NI support, VT-x support, greater cache, etc.

Bring both up on ark.intel.com and the only differences are clock speed (300 MHz), cache (1MB), and VT-d support.

Captain Ned wrote:

Aether wrote:

I am willing to spend more for a Core i3 to get the AES-NI support, VT-x support, greater cache, etc.

Bring both up on ark.intel.com and the only differences are clock speed (300 MHz), cache (1MB), and VT-d support.

Is Intel going to offer H210, B250, and H270 chipsets? I really haven't been keeping up with Kaby lake. Been keeping more track of Zen.

There was a rumor of an H270, but don't forget that existing 100-series boards will support Baby Lake with a firmware update.

With Haswell refresh there was only Z97 and H97. Lower-end PCHs like B85 and H81 weren't updated.

It bears mentioning that unlike VT-x, VT-d is not only something that needs to be supported by the motherboard as well, but also doesn't really do anything beneficial unless you have devices you want to exclusively make available to your VMs.

And some motherboards don't support it, or don't support it correctly. Some devices don't come with hardware features that allow you to make use of it, there are restrictions based on bus setups etc... And if you have just one disk/controller, one NIC, one GPU, etc..., it doesn't serve any practical purpose anyway.

---

Think of it like this:

VT-x, X for ISA eXtension: these are the instructions that make the actual CPU more virtualization friendly. It might need to be enabled in the BIOS, but it's not really dependent on anything else. Pretty much every Intel CPU, outside of maybe some Atoms, has supported this for a long time now.

VT-d, D for Device: This is support for virtualizing devices. It is an IOMMU, and unless the motherboard, the firmware, and the device you intend to virtualize in a VM all properly and correctly support it, it doesn't work. Most Intel CPUs support this, but until very recently Intel seems to have wanted to keep VT-d and ECC support separate on the lower end for segmentation reasons. Evidently this still applies on for some of the newer Pentiums, but not the newer i3s. Hard to keep track of it, as always, consult ARK.

---

tl;dr: VT-x is a win (sometimes even a requirement) for everyday hobbyist/tinkering VMs. VT-d is not. It is something you should only care about if you already know you need it and how you're going to use it. It is part of *FULL* stack decision, not a separate CPU feature.

Yup. VT-d generally doesn't work properly on consumer motherboards, even if your CPU supports it. You need server/workstation class hardware, and it only helps you if you are trying to set up VMs that have dedicated "bare metal" access to PCIe devices.

VT-x (or SVM in AMD-speak) is the one you care about if you run VMs at all.

Thanks guys for clearing this up. I really wish that Intel(and everyone else) would quit this product segmentation where it doesn't make any sense.

whm1974 wrote:

I really wish that Intel(and everyone else) would quit this product segmentation where it doesn't make any sense.

That's not really the problem honestly. The name is. They should have called it anything other than VT-d.

The segmentation like this of VT-d between the new Pentiums and new i3 isn't about enthusiasts, it must be about some OEM making something that could otherwise save a few bucks by using a Pentium. That's the only explanation that makes sense.