System Builder Marathon, June 2011: $500 Gaming PC

[pauldh]

I say you should ask Paul if he can get approval for a separate article using those parts.

Ah man, too many commitments as is; I'm swamped until the next SBM rolls around (and beyond)!

Go for it if you have time. I Don't really see the point though; these things are quarterly now. Any follow-up article shouldn't be able to capatilize on CPU and GPU prices we didn't have access to in our SBM; That's cheating. Sub in a 6850, we could have done it at budget. Test both GPU's...even better data.

I'll say this, it is a nice system and he is pretty close to spot on where I'd like to go for the next SBM, all depending what comes down the pipe of course. Wouldn't mind changing up the mobo and PSU. 6870 pricing better stay the same, or I need more budget for more GPU. The "gaming" system should be targeted at the now-sensible 1920x1080.

Enjoy your weekend!

 

[AppleBlowsDonkeyBalls]

[citation][nom]striker410[/nom]Not to continue to come down on you guys, but I keep seeing "in any task that isn't gaming". Do me a favor, read the article name. Yes, that's right, it says gaming PC. Just making sure you caught that.[/citation]

The only problem is the fact that if you trade off the Phenom II X4 955 for the Core i3 2100, you won't be able to get a Radeon HD 6870 into the budget. And having a Radeon HD 6870 with a Phenom II X4 955 will be better for gaming than having a Core i3 2100 with the Radeon HD 6850, plus the Phenom II X4 955 is faster in anything decently threaded. What you end up with is a rig that is faster at everything other than audio encoding, which takes almost nothing on either anyway.

To Paul:

You should use these parts for the next $500 SBM Gaming PC three months from now if prices stay the same, which I think they probably will. That will bring never before seen performance for the price since the 955 at a 3.5-3.6GHz OC will be faster than the 925 at 3.2GHz, plus the Radeon HD 6870 should be quite a bit faster than the Radeon HD 6850.

 

[pauldh]

^ We think alike. If building now, I do agree the 955BE is steller for the buck, hence it's spotlight in both the intro and conclusion.

The X4 925 was OC'ed to 3.4GHz (not 3.2 GHz), so we won't get too much more from the 955BE. While the BE boxed cooler is better, that last 925 did quite well at low voltage. I doubt we'd get to boost voltage and maintain acceptable temps, so we'd have to see just how the chip scales. It varies; My 955BE will not do 3.6GHz without a couple bumps in voltage, and the retail cooler just won't cope with that.

Will most definately look for a bump in GPU if possible. Keep in mind, while you'll see the next SBM in three months, we need to order components well before that time.

 

[r0aringdrag0n]

http://www.anandtech.com/bench/Product/289?vs=88

 

[Onus]

[citation][nom]r0aringdrag0n[/nom]http://www.anandtech.com/bench/Product/289?vs=88[/citation]
We already know that. My concern, and I believe it is shared, is that a more realistic real-life test that includes background tasks present in other software used by gamers, while gaming (e.g. a VOIP client, maybe a browser window, anti-virus) will bog down the hyperthreaded dual-core Core-i3 much more than it will a true quad, even if it is a slower AMD chip.

 

[eddieroolz]

I for one don't like this machine. Sure, it may get better frames at lower power consumption - but the limitations imposed by the processor and the chipset is just to big to ignore.

For example, having only one PCIe slot would've been acceptable in 2007; but in 2011, simply for extensibility's sake I would want at least two.

 

[humble dexter]

h61:55+
p67:104+
am3:38+
am3+crossfire:45+
990fx:129+
990x:110+
970:98+

I3:100
I5:174
955x2:75
965x4:116


There's a difference between a cheap motherboard, and a locked motherboard. And the H61 chipset isn't just a cheap P67 motherboard : It's a sub-P67 locked motherboard !

So it was a choice between :
- A bit more frame rates on yesterday's benchmarks today, using a locked dual-core I3, on a cheap BUT locked H61 motherboard that can't be upgraded to a high end K series when the time is right.
- A bit more frame rates on tomorrow's games using an unlocked Phenom II, on a cheap and unlocked AM3+ motherboard that CAN be upgraded to a high end Bulldozer when the time is right.

What I would choose ?
- For writing a review in Tom's Hardware, I would be more concerned about achieving the best performance on (yesterday's) benchmarks for the sake of my review, then I would care about the H61 chipset being locked, games shifting from 2+ cores to 4+ cores, or whatever... since I would be giving that PC to a complete stranger just after my review was done.
- If I was building a gaming PC for myself (or someone I happen to care about), for a 500$ build I would go for the overclocked Phenom II on an AMD3+ socket, with just a bit less performance then I3 to expect on benchmarks from previous games, and a bit more performance then I3 to expect on everything else, including on tomorrows more multi-core demanding games, and the ability to upgrade the configuration at a later date just by switching to a high end unlocked CPU (without the hassle or cost of a new motherboard).

My opinion :
The reviewer made the right choice for his review (specially if he was aiming to squeeze a Sandy Bridge in a 500$ build the best he could), and has the benchmarks to prove it... but I wouldn't recommend using his advice if you're creating a 500$ build for someone you care, instead of writing a review.

My suggestion to Tom's Hardware :
Please write an article on "I3-2100 vs Phenom II in 4+ core games", to provide us the insight we lack on the best kind of future-proof CPU we should be buying for a value oriented PC today, to meet next year's gaming requirements.

 

[pauldh]

[citation][nom]jtt283[/nom]We already know that. My concern, and I believe it is shared, is that a more realistic real-life test that includes background tasks present in other software used by gamers, while gaming (e.g. a VOIP client, maybe a browser window, anti-virus) will bog down the hyperthreaded dual-core Core-i3 much more than it will a true quad, even if it is a slower AMD chip.[/citation]
I've maintained at least one (seperate) pure gaming system for 10+ years now, and it's always kept pretty clean for just these reasons.

I doubt we'd introduce such variables into the SBM, but what you are asking is well worth a look! I'm swamped, but will pass this idea along to the team and see if anyone would tackle a story focused on clean/average loaded system, comparing gaming CPU's. Personally I do not think the Core-i3 will be bogged down much but would love a chance to scour that data myself.

 

[Onus]

Thanks paul. That's the most I'd have any right to expect.

 

[striker410]

[citation][nom]Humble Dexter[/nom]h61:55+p67:104+am3:38+am3+crossfire:45+990fx:129+990x:110+970:98+I3:100I5:174955x2:75965x4:116[/citation]
if that's what you think mobo pricing is, you might wanna go take a look at newegg.

I agree an AM3+ mobo would be nice, but they'd have had to take a hit somewhere else. Newegg is down right now, but in previous searches I've been unable to find one that's less than $80. With crossfire and usb3? $100. That would have required them to take a hit in the graphics, which is the bottleneck anyway. They could not have done that. They would have to go just AM3.

Hopefully prices will come down in the future, and in 3 months we can have a much better system

 

[humble dexter]

[citation][nom]jtt283[/nom]gaming (e.g. a VOIP client, maybe a browser window, anti-virus) will bog down the hyperthreaded dual-core Core-i3 much more than it will a true quad, even if it is a slower AMD chip.[/citation]
The general idea is : What percentage of the full CPU is being used in your gaming results ?
Because whatever percentage of the CPU is left idle during the game benchmark, is what's available for more background tasks today, or for better threaded games tomorrow.

 

[humble dexter]

[citation][nom]striker410[/nom]if that's what you think mobo pricing is, you might wanna go take a look at newegg.[/citation]
I copy-pasted those numbers in my previous message by mistake, sorry about that ^^

Those are in fact real numbers, without any special rebate, but they are all expressed in Euros.
Also the 45 Euro AM3+Crossfire is a "Crossfire + AM3 socket", and not a "AM3+ + Crossfire" motherboard, meaning you won't be able to upgrade it with a bulldozer, but you can upgrade it with a second graphics card instead.

With Phenom II and AM3 prices being pushed further and down by Sandy Bridge, upcoming Bulldozer, and the new AM3+ socket, with gaming bottleneck shifting from CPU to GPU lately (unless you're willing to drop graphics setting to lower levels), and with entry level Bulldozer rumored at ~190$, maybe providing the space for a second graphics card upgrade on a AM3 motherboard, makes a whole lot more sense then providing the space for a Bulldozer CPU upgrade, specially on a 500$ AMD build ?

Of course a crossfire an AM3+ motherboard would provide room for both a graphics card and Bulldozer upgrade (in case AMD ends up introducing a mainsteam dual core Bulldozer or something), but there are no cheap crossfire AM3+ motherboards available at this time : Only cheap crossfire AM3 motherboards.

 

[Onus]

With apparently no silicon changes between 800- and 900-series chipsets, I'd be concerned about any AMD board showing some kind of bottleneck when multi-GPUs are used. There's supposed to be some additional power control for Bulldozer, but I don't see how that might affect chipset performance; or was it the CPU alone causing the bottleneck? A study of multi-GPU set ups with a Bulldozer vs. a Phenom II CPU would be interesting.

 

[IbrahimV]

Paul, when making builds, do you only consider lowest price or do you also put into consideration getting parts with rebates, coupons or cash back offers?

 

[jgillispie]

Personally if my budget only allowed a $500 system I wouldn't be OCing. I couldn't afford the risk of damaging components. Not to mention going with an 1155 chipset over a dying AM3 could allow me to vastly upgrade the machine later on with an i7.

 

[mapesdhs]

Heh, that's why I mentioned the 540. Btw, this week I bagged an i7 875K, an
Asrock P55 Deluxe and 4GB Corsair DR3/1600 RAM on eBay for 160 UKP, plus a
Thermaltake Toughpower 1kW PSU for 64. Meanwhile, an i3 540 last week cost
me 53 UKP, a Gigabyte GA-H55M-S2H was 30 and a Thermaltake Toughpower 750W was
50. Sometimes 2nd-hand is such fun. I build PCs for friends/family, so I
acquire decent items as & when I can, though mbds can be a minefield on eBay.

Ian.

 

[lott11]

The reason for my out outlandish Roderic is simple.
Most people that would build a PC for a price point of $500, are not expecting to keep it like that for long.
No in a months time they would be upgrading it.
With the Intel how much is the expandability, what more CPU, more ram, faster drive, then what.
With the AMD you can do all of the above and more.
Yes Hyper Threading is all most the norm this days.
But Cores do all most the same, at this price range.
At the present time Intel Core i5-2400 Sandy Bridge 3.1GHz cost $189.
For the same price with AMD AMD Phenom II X6 1100T Black Edition Thuban 3.3GHz, or AMD Phenom II X4 980 Black Edition Deneb 3.7GHz.
This would be the most logical upgrades that a gamer on budget would do in 3 months time if he could.
Hyper Threading relies on ram and hard drive space speed.
To quote someone els,
“How does this usually work out? It's usually somewhere in the middle.
I could continue to say that simultaneous multithreading is what developers do once the hardware is already there, and that HT is called Symmetric multithreading... but is anyone else going to remember that? Yeah... probably not.” Quote By Mime Troll Hunter.
And it is all a numbers game, and who could take it the furthest for less cash and for the longest time.

 

[mapesdhs]

[citation][nom]lott11[/nom]... Hyper Threading relies on ram and hard drive space speed. ...[/citation]

Really? Where/what is your evidence for this? Got any URLs you can point to? ie. impact of RAM bandwidth/MHz/latency on HT performance for a range
of tasks? And is that for a CPU at a fixed clock with no Turbo, etc., for
for a CPU running with everything at default so it can Turbo up when viable?
And all of this with various types of disks/storage? Mechanical, SSD, etc.
That's a huge number of variables you're claiming to be certain of. Where's
your data?

Ian.

 

[lott11]

I new that if I made some dome remarks, it would generate some answers.
I have been purchasing parts from the egg for the past 12 years.
And when I build a PC for someone, it has to work for minimum of 3 – 5 years.
And it has to be upgraded at least 1 time with in 2 years, so it has to upgrade at list 30 percent.
And that build just did not make any of those standers.
And it looked like some thing that Hp would build, with little to no upgrade potential.

 

[lott11]

Really? Where/what is your evidence for this? Got any URLs you can point to? ie. impact of RAM bandwidth/MHz/latency on HT performance for a range
of tasks? And is that for a CPU at a fixed clock with no Turbo, etc., for
for a CPU running with everything at default so it can Turbo up when viable?
And all of this with various types of disks/storage? Mechanical, SSD, etc.
That's a huge number of variables you're claiming to be certain of. Where's
your data?

Ian.


this are just some the links, others can't be shown do to confidentiality, but just Google it and see other reports.
And ask why is the cache on a Intel L3 greater then on a AMD.
Also see on number crunching or data transfers to GPU.
And yes the bottle net is the GPU and data true put.
The way that AMD dose it is different to Intel, Cores on a Intel have more time cycles.
AMD data true put flow is directed different it uses less cache.
http://technet.microsoft.com/en-us/library/cc298801.aspx
download.intel.com/design/intarch/papers/30349001.pdf
http://www.matrixscience.com/help/pc_specs.html

 

[hangfirew8]

[citation][nom]dread_cthulhu[/nom]Personally, I'd be a bit wary about a 380W PSU pushing a 6850... I think the manufacturer recommends a minimum of 400W on that one. Not including all the other components asking for power in that system. I always throw in bigger unit than I calculate anyway, so this is kinda scary for me.[/citation]

Installing a grossly oversized PSU is only necessary when using a grossly overrated cheap PSU. A quality PSU (like Antec) can handle the load, and the graphs show it.

Throwing in a bigger unit is great for future expandability, but easy expandability is one of the first things you give up when you have a tight budget like $500. It hurts less to toss parts, though, when they cost as little as the Antec (and you can probably find a good use for it in some pal's eMachine that just blew smoke).

 

[mapesdhs]

lott11 writes:
> Ian.this are just some the links, others can't be shown do to

Neither of those links has any data that backs up your claim. As I
suspected, you have no real references to justify what you said.


> confidentiality, ...

Nonsense. If such a thing was true, it would be easy to benchmark
and demonstrate, there's no reason why any such test results would
be confidential. Surely you've run some tests yourself?


> ... but just Google it and see other reports. ...

You're the one making the claim, why should I do the work?


> ... And ask why
> is the cache on a Intel L3 greater then on a AMD. ...

Ask the reverse: why is the cache on AMDs smaller? I'll answer it for
you: because if it was larger they wouldn't be price competitive for
the performance offered. Intel has it partly because it can and partly
because it makes sense for the design being used, and btw remember the
cache sizes on Intel chips dropped quite a lot with the introduction of
Nehalem (have you forgotten the huge 12MB L2 used by the top-end
Core2Quad and Core2Extreme series? A large proportion of the entire
Core2 range used 3MB L2 per core - Nehalem reduced this by a 3rd).


> on a Intel have more time cycles.AMD data true put flow is directed

None of that answers my question or proves your point.

You said HT is dependent on RAM speed and disk I/O. I'm simply asking
for your rationale/data for this claim. So which of the tests run by
Paul would be affected, in what way and why? Pointing to obscure
articles on microsoft.com or other tech sites doesn't prove anything,
especially given the $500 is for tasks completely unrelated to those
references. Plus, real server tasks are bottlenecked far more by
system I/O, or as John Mashey once put it, "It's the bandwidth, stupid!".

So far the only conclusive HT effects I've been able to show for gaming
is that it doesn't help for older games, can/might/does help for newer
games, and that reducing RAM speed from 2GHz to 1600MHz reduces fps
rates by barely more than 1% (much less than the speedup offered by HT
for those situations where it can help, especially video encoding, etc.)

Ian.

 

[lott11]

I did even have to proof it, just look at this article http://www.tomshardware.com/reviews/amd-a8-3850-llano,2975.html
yes it is a bought the the lano and it shoes the point of ram and it's impact, its same with Intel of cycles in Hyper technology and it's recommendations for mobo development.
AMD are more relax, you have to read all recommendations to see all it potential.
And I may be a newbie in this forum but not in the industry.
It is a balance of components that get you to the top.,
and I worked with IBM,Hyundai,Mitsubishi,Motorola,Intel, TI, so it is a balance.
What works with what, and where.

 

[mapesdhs]

There's nothing in that article that backs up your claim. I've asked twice for data that
proves your specific claim about HT being dependent on RAM/disk speed, but so far zip.

Maybe your claim is infact true, but saying it is without a proper rationale makes no sense.
And if it's something you're so sure of, why not run some tests yourself?

Ian.

 

[Onus]

Lott11, I generally have to agree with mapesdsh that you have not made your points at all, but THIS:
[citation][nom]lott11[/nom]...And when I build a PC for someone, it has to work for minimum of 3 – 5 years. And it has to be upgraded at least 1 time with in 2 years, so it has to upgrade at list 30 percent.And that build just did not make any of those standers.And it looked like some thing that Hp would build, with little to no upgrade potential.[/citation]
...I believe to be absolutely true. When I build a PC, I never build something that has zero upgrade potential or otherwise won't last. This month's "$500" (it went over) build can be upgraded with a faster CPU and/or GPU, but sooner or later will hit platform limitations of the cheap mobo that was used. It was a one-off to win benchmarks, not something I (or you, apparently) would build; but that's ok, I don't think these articles are intended to be buying guides (although that sometimes happens). That doesn't mean they're useless, on the contrary I think the discussion of this one clearly shows its limitations and that it is ultimately a dead end. Anyone who's been on the forums for even a week or two has had the folly of buying a cheap PSU preached at him (and I'll admit to being one of the preachers), and benchmarks readily allow CPUs and GPUs to be compared, but this article (and the discussion that followed) showed the problems of buying a cheap mobo.