Dear forum members,
We are delighted to inform you that our forums are back online. Every single topic and post are now on their places and everything works just as before, only better. Welcome back!


Discussion on Article:
May Power Be with You: Find the Best CPU for a Single ATI Radeon HD 5870 and ATI Radeon HD 5870 CrossFireX Configuration

Started by: TAViX | Date 12/10/09 10:29:15 PM
Comments: 29 | Last Comment:  02/08/10 01:13:41 PM

Expand all threads | Collapse all threads


You should have use at least one or two Core 2 Quads. Comparing a Core 2 Duo with i7, is just comparing apples with tires.
0 0 [Posted by: TAViX  | Date: 12/10/09 10:29:15 PM]
- collapse thread

Oh well, it's extremely time-consuming, and perhaps Xbitlabs did not have a Core2 Quad on hand at the time.

The performance of a Core 2 Quad is most likely going to be right in between a PhenomII and a Core i7 in this review, if this helps.
0 0 [Posted by: Bo_Fox  | Date: 12/12/09 11:55:17 PM]

Thank you for this article. I have always needed this comparison. Your conclusion also highlights some of the differences between dual and quad core.

I now know that my aging 9550 will do just fine overclocked to 3.6 Ghz for the coming years and no need for an upgrade to i7 just yet. I might only jump when the i9 or later comes out.

@TAViX, I don't think the point of this article was to compare the CPU's but rather to give you an indication if gaming would be affected by the type of CPU you have. The point derived from this article is that you are still gaming capable with a dual core just over clock it a bit and you are set for the next 12 to 18 months

Great article!
0 0 [Posted by: Nictron  | Date: 12/10/09 10:47:27 PM]
- collapse thread

Agree, but it would have been nice also to see the the performance of a Quad, like 9550 or 9650 compared to i7 for example...
0 0 [Posted by: TAViX  | Date: 12/11/09 03:41:33 AM]

A decent review between core 2 duo and i7, but all your AMD data needs to be thrown out and redone properly. Seriously how can you have test systems trying to be equal yet the AMD system gets only 8x lanes when in XFIRE mode with the 790GX chipset. The numbers for single cards are likely fine but for Xfire, they are so skewed by being stuck with only 8x bandwidth.

I also dislike the fact that you used 2gig for the core2 and the AMD setup, yet used 6gig for the i7, you could have at least put in 4gig in the core2 and i7, or took the i7 down to 3gig. Im sure though the reviewer tried to stack things as much in favor of the i7 system as it possibly could.

Usually Xbit labs has some unbiased reviews, but year after year i keep seeing reviews that try to be unbaised in the content, but are clearly biased in the test setup with the parts being used.
0 0 [Posted by: daseto  | Date: 12/11/09 02:58:12 PM]
- collapse thread

The numbers for single cards are likely fine but for Xfire, they are so skewed by being stuck with only 8x bandwidth.

If it occurred, the CrossFireX on AMD platform in hard graphics settings and in 1920x1200 the results will be low then Intel platforms. And it does not occur.

I also dislike the fact that you used 2gig for the core2 and the AMD setup,

This is not right! 4 GB on AMD and 4 GB on C2D. See the screenshots.
0 0 [Posted by: Jordan  | Date: 12/11/09 10:31:24 PM]
Jordan did you miss the testbed and configuration page? I could care less about screen shots as they dont mean anything, however, what the reviewer says he used does matter.

System memory:

* DDR2 SDRAM 2 x 1 GB Corsair Dominator TWIN2X2048-9136C5D (1142 MHz / 5-5-5-18 / 2.1 V);
* DDR2 SDRAM 2 x 1 GB CSX DIABLO CSXO-XAC-1200-2GB-KIT (1200 MHz / 5-5-5-16 / 2.4 V);
* DDR3 PC3-12800 3 x 2GB OCZ Platinum Low-Voltage Triple Channel (Spec: 1600 MHz / 7-7-7-24 / 1.65 V);

Notice the 2x1gig kits for the C2D and AMD setups.

As for your stuff about Xfire, you cannot compare apples and oranges and expect results that you can actually use. The intel setups had 16x pcie lanes in single and in xfire mode, where the AMD board had 16x in single and 8x in xfire. The most powerful damn gfx cards are always sucking as much bandwidth as possible, and given that the AMD setup had only half the bandwidth available compared to the intel systems means xbit labs was comparing apples and oranges yet trying to say they are apples to apples.

Xbit labs screwed up this review horribly as it is not anywhere close to an unbaiased review with everything i have mentioned already.

0 0 [Posted by: daseto  | Date: 12/12/09 11:46:00 AM]
In AMD platforms and in Core 2 Duo was used 4 memory modules:
* DDR2 SDRAM 2 x 1 GB Corsair Dominator TWIN2X2048-9136C5D (1142 MHz / 5-5-5-18 / 2.1 V);
* DDR2 SDRAM 2 x 1 GB CSX DIABLO CSXO-XAC-1200-2GB-KIT (1200 MHz / 5-5-5-16 / 2.4 V);

Read the text attentively next time:
Despite my using fast DDR2 modules in the system, the integrated memory controller of the AMD processor imposed some limitations resulting in a performance reduction. I could not use CAS Latency 4, Command Rate 1T and a step-up memory divisor. Unfortunately, I did not have two good 2GB modules and it was not right to sacrifice 2GB of RAM (by removing two modules) to benchmark modern games in Windows 7.

And only after that start discussion.

About PCI-E 8x VS PCI-E 16x go to this article:
and try to see a difference.
0 0 [Posted by: Jordan  | Date: 12/12/09 01:31:20 PM]
its still not clear what you put on the review

you used 4x1gb of ram 2 different makes was used (even when i read the review it looks like you Did use 2gb of ram in both systems) most reviewers would not use 2 different brands of ram

just noticed {In AMD platforms and in Core 2 Duo was used 4 memory modules:} did not see that before, even i missed that part

this part Needs re wording
I did not have two good 2GB modules and it was not right to sacrifice 2GB of RAM (by removing two modules) to benchmark modern games in Windows 7.

at least in there put in that 4gb was used in both Intel and AMD (due to the fact you used 2 dif brands of ram some reviewers norm do this when one set of ram does not work on intel side or AMD side so listing 2 types of ram that are 2x1gb each i would of asumed that it was 2gb of ram each system as no one would norm use 4gb of ram in 1gb sticks more so in an review)

just clean up that part or they can let them read this part any way so they can work it out (i edit my other post as i even thought it was 2gb for core2 and amd)
0 0 [Posted by: leexgx  | Date: 12/12/09 09:31:21 PM]
These are PCIe 2.0 lanes. The difference between x8 and x16 will not involve any CPU bottlenecking, which is the point of the review.

2GB vs 4GB RAM will have nothing to do with framerate either. You are lacking a lot of comprehension of PC hardware it seems.
0 0 [Posted by: JonMCC33  | Date: 12/15/09 06:02:40 AM]

plz fit amd phenom II x4 fit on a x16 x16 790fx board if making benchmark on crossfireX configuration... it seem unfair for x8x8 vs x16x16 in x58 and x48 board...
0 0 [Posted by: secret_tidus  | Date: 12/12/09 02:27:40 AM]
- collapse thread

Look at the link that Jordan posted. It shows that the difference between 8x and 16x PCI-E lanes for two 5870's is usually less than 1-2% in most games.
0 0 [Posted by: Bo_Fox  | Date: 12/12/09 11:58:33 PM]

This article was a breath of fresh air.

Great article.
0 0 [Posted by: oldDummy  | Date: 12/12/09 04:06:41 AM]

Would be nice to see you use a Phenom II X2 in these tests too, even if nothing but for a less Intel oriented set of tests.

Would have been nice to use DDR3 on the Phenom system too, you could have used the Gigabyte GA-MA790FXT-UD5P which has two PCIe x 16 slots and supports DDR3.
0 0 [Posted by: GavinT  | Date: 12/12/09 04:29:09 AM]
- collapse thread

You would have just been more upset, from seeing the phenom x2 getting smoked by the c2duo's.
0 0 [Posted by: actionjksn  | Date: 01/25/10 06:27:30 PM]

Wow .. I guess that was lots of work .

So I am sorry to demand some retest ^^

The i7 with DDR3-1600 are running their L3 Caches with 3.2 GHz.

I would like to see some L3 / NB / Uncore OC with AMD, too.

Or are you limited by the board ?

I assume L3 for AMD was running at 2.27 GHz, because you raised the HT reference clock to 227 Mhz.

Well maybe save a re-test until you get a decent 790FX board with DDR3 ;-)
0 0 [Posted by: Bingle  | Date: 12/12/09 04:29:49 AM]

This article tells a different story compared to the one written

Unlike Xbitslabs. The phenom platform was tested fairly using 16x instead of 8x for the video card, and 4 gigs of ram was installed instead of 2 that Xbit had installed.

The Phenom machines were able to match the i7, and even beaten it in some games.
0 0 [Posted by: xenocea  | Date: 12/12/09 09:00:49 AM]
- collapse thread

No, Xbitlabs used 4GB of memory for Core2 and PhenomII, not 2GB. Also, the difference between x16 and x8 PCI-E is usually less than 1-2% in most games, even for two 5870's in crossfire.
0 0 [Posted by: Bo_Fox  | Date: 12/12/09 11:48:53 PM]

It would be nice to see a ddr3 am3 system and a core 2 quad added to these tests

as for a few of the comments posted here
there is reviews that show 8x8x pci-e is fine for 5870
but it would be nice to see xbit do a comparison using 16x 8x & 4x crossfire\sli with 5870\5970 and whatever nvidia comes up with next
unfortunately getting the hardware is always going to be a problem

I think they used 4x1g for the core 2 and amd system while not as bad as running 2x1g of ram it may not as good as running 2x2g when overclocking
0 0 [Posted by: dasa  | Date: 12/12/09 04:04:10 PM]
- collapse thread

At least the DDR2-800 had CAS latencies of 5-5-5, which is not too shabby. I do not think it would be too bottlenecking for a Phenom II, which still did very well for a single 5870.

A core 2 quad would be right in between a PhenomII and a Core i7. Just trust me on this.
0 0 [Posted by: Bo_Fox  | Date: 12/12/09 11:53:40 PM]

Good review but please use a proper Amd Mobo with enough lanes + RAM, its not fair atm.
0 0 [Posted by: 3Dkiller  | Date: 12/12/09 08:23:48 PM]
- collapse thread

It's "hardly not fair". "Hardly", since the difference is usually less than 1-2% in most games, even for two 5870's in crossfire when switching from x8 to x16 PCI-E lanes.
0 0 [Posted by: Bo_Fox  | Date: 12/12/09 11:51:46 PM]

before reading this review, i have to agree an gaming system or users who are going to be gaming would be using 2x2GB (i use 8gb) of ram for dual channel setups (AMD/Core2/i5) and 3x6gb for I7 (dual Chan over triple Chan does not affect game performance any way in some cases triple harms it but your talking close to margin of error numbers here)

but overall it should not matter if windows 7 was used, the systems should of not ran out of ram

X8 over X16 PCI-e 2.0 slots should have no affect on the results as well (i5 is 8x/8x in CF or SLI due to the CPU providing limited lanes) you not likely going to saturate an PCI-e at 8x in 2.0 mode

it may have been that this review was relating to users who currently have 2gb of ram in there systems but choice to upgrade the video card as a lot do that, any one with an I7 will most likely have 6gb of ram installed as its only less then an year old

now i have made my post i read the review (TAViX comment was daft post to not understanding why this review) and comment on it again

(due to the way this review was done and worded he used 2 dif brands of ram thats 4x1gb he used in the core2 and AMD, he may have updated it or not, just an foot note for any one els who thinks he used 2gb of ram in the core2 or AMD system when in fact it was 4gb)

i have looked at it seems to gone the way i expected it the I7 is 15-30% faster then an Phenom II clocked at the same speed this is know fact, i would used 4gb of ram but it would have Not affected the end result any way as these are Clean systems, the CPU overclock would have made more of an result

if i can just get some one to buy my QX6850, mobo, heatsink and ram i would get an I7 now clocked at 4ghz but £600 to much until i can sell my parts, i use AMD Athlon II systems for all my norm system builds {Phenom I/II run to hot for norm users as to why i use Athlon II, if AMD had not made them i would of switched to intel as there lower end 45nm run cool, and amd had removed all there lower clocked dual core X2 from sale at the time} amd have Far better motherboards/barebone systems then intel at the low end
0 0 [Posted by: leexgx  | Date: 12/12/09 08:49:19 PM]

Wow, a superb article!!!!!!!!!!!!!!!!!!!!!!!
0 0 [Posted by: Bo_Fox  | Date: 12/12/09 11:42:55 PM]

An Intel Q9*** CPU included in the test would have been really interesting clocked @ the same frequencies but then i7 would not look like such a good upgrade path (which it is NOT)

the x8 vs x16 lane argument is a non issue, they are very close but 2x2GB usually has lower latencies than 4x1GB so maybe borrow a decent kit next time
0 0 [Posted by: alpha0ne  | Date: 12/13/09 01:29:41 AM]

One great of an article, welldone Sergey.

(Any chance, the page layout of this site can be modified; the main article is placed too much to the right, this makes reading a bit uncomfortable)

0 0 [Posted by: Ibnsina  | Date: 12/13/09 02:46:38 AM]

what's the meaning of" There is almost no performance growth on the weak Intel Core 2 Duo. "?
almost no performance growth ?
0 0 [Posted by: realice  | Date: 12/14/09 02:41:53 AM]

For both INTEL test systems you are using motherboards with two pcie x16 lines
and for AMD you are using board with integrated graphics and one x16 line and one x8 line.
This test is just not objective!!!
So it s not a test
0 0 [Posted by: behemoth  | Date: 12/15/09 12:44:01 PM]

Poor test.

As most people have already mentioned the shamefully obvious hardware stuff, I'd like to point out that the overclock on the AMD system is poorly done.

I stopped reading when I saw the HT Link speed of 2200Mhz and no mention at all about the CPU-NB frequency on the AMD system. Overclocking is not setting a few sliders to the right. It's real world performance that this audience wants, not just impressive settings.

The Phenom's performance increases dramatically by upping the CPU-NB frequency (That's the integrated memory controller). At 4.0 Ghz cpu clock, you'll want to have the CPU-NB at 2600Mhz to 3000Mhz. Looking at those Everest results, rest assured the CPU-NB is not working at those speeds.

Secondly, you want to keep the HT link speed at 2000Mhz (stock). Clocking it higher will seriously damage your 3D performance. With dual card setups however, a small overclock on the HT link might increase performance.

Charts? Here you go:

Now ofcourse this info is out of scope for this article, but if you are showing overclock results at least make it a proper overclock and use proper/similar hardware.
0 0 [Posted by: johngoodman  | Date: 02/08/10 01:13:41 PM]


Back to the Article

Add your Comment