Less than spectacular retail X6800 3D results

[lopri]

Someone @XS received his X6800 yesterday and started his experiment. I've been following the thread and the results he got seem to be a somewhat mixed bag, as far as 3D gaming is concerned.

Of course we know the high res 3D gaming is mostly GPU-bound. But at the same time we often hear how those SLI/CrossFire setups are limited by CPUs. My expectation WRT Conroe was so high that we'll finally see some dents in the benchmark graphs (instead of all lined up bars regardless of clock speeds) w/ high res, high quality setting.

http://img167.imageshack.us/my.php?image=3dmark06run35js.jpg

The above screenshot is the most he could get so far with his X6800. He also compared the result with that of his X2 rig.

Core 2 Duo @2.93GHz
3DMark Score 10230 3DMarks
SM 2.0 Score 4783 Marks
SM 3.0 Score 4417 Marks
CPU Score 2517 Marks

Detailed Test Results

Graphics Tests
1 - Return to Proxycon 38.177 FPS
2 - Firefly Forest 41.541 FPS

CPU Tests
CPU1 - Red Valley 0.795 FPS
CPU2 - Red Valley 1.275 FPS

HDR Tests
1 - Canyon Flight (SM 3.0) 39.57 FPS
2 - Deep Freeze (SM 3.0) 48.765 FPS

A64 X2 @2.8GHz
3DMark Score 9252 3DMarks
SM 2.0 Score 4357 Marks
SM 3.0 Score 4288 Marks
CPU Score 2039 Marks

Detailed Test Results

Graphics Tests
1 - Return to Proxycon 35.321 FPS
2 - Firefly Forest 37.299 FPS

CPU Tests
CPU1 - Red Valley 0.65 FPS
CPU2 - Red Valley 1.023 FPS

HDR Tests
1 - Canyon Flight (SM 3.0) 39.872 FPS
2 - Deep Freeze (SM 3.0) 45.895 FPS

Both with stock 7900GTX SLI (650/800). Now here is my personal record (ORB link in the sig)

A64 X2 @2.97GHz
3DMark Score 10314 3DMarks
SM 2.0 Score 4875 Marks
SM 3.0 Score 4801 Marks
CPU Score 2249 Marks

Detailed Test Results

Graphics Tests
1 - Return to Proxycon 39.767 FPS
2 - Firefly Forest 41.49 FPS

CPU Tests
CPU1 - Red Valley 0.718 FPS
CPU2 - Red Valley 1.127 FPS

HDR Tests
1 - Canyon Flight (SM 3.0) 45.001 FPS
2 - Deep Freeze (SM 3.0) 51.019 FPS

My GTXs were OC'ed to 700/900.

What this tells me is we still won't see any improvement in high-res, high-quality gaming w/ Conroe. By high-res, high-quality I mean something like 1920x1200, and/or dual graphics cards setup. Assuming there will be faster graphics card coming out within next 6 months, it seems like Conroe's effect on gaming will be minimal. We'll still be GPU-bound. It looks like OC'ed X2 will have no less than 90% performance of Conroe in gaming.

Another worrying thing is that the original poster @XS wasn't able to stably run the X6800 higher than 3.2GHz. He could boot @3.4GHz but it wasn't stable enough to run 3D. He also says the load temp @>3.0GHz was almost 70C(!) and we can see it from the screenshot above. He later said he was able to lower the load temp to 60C, using all-copper heatsink with a 5000RPM(!) fan. But how much more MHz he was able to gain with heavier cooling is unclear.

This is a totally different picture from what I was looking forward to. I've seen many ES Conroes running over 3.5GHz @stock Vcore and thought retail versions would do even better. And thought we'd finally have a CPU that'll actually push GPUs harder than 3200+ at high resolution. (you know, how many times have wee seen 1600x1200/4AA/8AF benches where 3200+ and FX60 yields same numbers?) Note we're talking about X6800 here, not E6600.

Were those ES Conroes all cherries? I'm somewhat concerned.

P.S. I wrote this purely from 3D gaming and OC'ing point of views. I know Conroes will excel all the other areas.

 

[TuxDave]

I'm just a little confused how a 6% increase in clockspeed of the X2 resulted in > 10% performance increase.

Edit oh.. nevermind, the 3rd set of benchmarks had an overclocked video card. Can you your scores at stock GTX speeds for better cpu comparison?

 

[freethrowtommy]

If this is the case... AMD is still looking pretty darn good.

If you look at the CPU tests strictly... you can see a nice increase with the Conroe tho. That may be the benchmark you want to look at when comparing the X2 to the Conroe.

 

[avi85]

the conroe is a boiling hot chip!!!
look at that temp in speedfan, 69 degrees celsius!!!
and it's not even overclocked in that pic!!!
it's like the smithfield all over again...

 

[coldpower27]

Oh please, comparing a overclocked GPU vs a stock 1.... 3D Mark 2006 is GPU bound.... 1000 Point with the same GPU subsystem is failry impressive.

 

[theprodigalrebel]

Comparing the CPU scores for 3DMARK,
Conroe @ 2.93GHz: 2517 Marks
X2 @ 2.8GHz: 2039 Marks

~23% increase for the Conroe. Isn't this kinda good? Or does this test, like all synthetic tests, reveal nothing concrete about real-world performance?

 

[dmens]

Originally posted by: avi85
the conroe is a boiling hot chip!!!
look at that temp in speedfan, 69 degrees celsius!!!
and it's not even overclocked in that pic!!!
it's like the smithfield all over again...

what's the cooling setup? 3ghz conroe at load draws less power than a 3ghz presler idling.

 

[broly8877]

Faulty temp sensors, guys.

Almost all boards are showing 10-25c more than what they actually are.

Remember, Intel says the max the chip can handle is ~65c, so that reading is obviously not right.

 

[Marmion]

I've read through that thread and many other threads there at XS, and it is a general consensus that the Bad Axe and Asus mobos over-report temps by 15-20 Celcius.

Also, the guy in that thread was only playing with the multi's, not FSB, so he managed to raise the multiplier on stock volts from 11x to 13x, although not stable, on stock cooler, and volts for everything.
He was also running on stock volts for the RAM, which he has admitted that rasing the volts on the ram, he could get it stable at the higher overclocks.

So, hold your horses and wait untill the guy has had it for a couple of days and started overclocking the FSB and upping the volts

 

[Avalon]

I think you were expecting too much in high res gaming. It should have a nice impact in lower res gaming (10x7), and somewhat of an impact at standard res gaming (12x9. 12x10, 16x12).

 

[anandtechrocks]

Originally posted by: avi85
the conroe is a boiling hot chip!!!
look at that temp in speedfan, 69 degrees celsius!!!
and it's not even overclocked in that pic!!!
it's like the smithfield all over again...

It's definately some issue with the motherboard. HKEPC did a Conroe review and used a temp probe. Here are their results:

http://hkepc.com/hwdb/x6800vsfx62-11.htm

 

[lopri]

I see an issue with their benchmarks.. 400+ FPS in BF2, @1600x1200/4AA? Do BF2 have a DX7 mode?

 

[TekDemon]

Give people some time to tweak it out, most of the people who were OC'ing the ES chips were overclocking pros, while the people getting ahold of the regular chips now probably aren't quite as skilled at being able to milk every last mhz out.

And obviously most newer benchmarks and games are GPU bound at the higher resolutions so the bump is CPU speed won't show a huge difference. But anything CPU intensive and CPU bound is going to show the big differences.

 

[n7]

If you run at very high resolutions, there isn't gonna be a huge difference between an X2 or C2D due to GPU limitations.

It's not rocket science.

 

[lopri]

Hate to quote HardOCP but they have now a front-page banner that says things in line with my observation.

Originally posted by: HardOCP
Real world gameplay analysis of Core 2 vs. Athlon 64 coming very soon. Shows different results than canned benchmarks? You betcha

 

[lopri]

Originally posted by: TuxDave
Edit oh.. nevermind, the 3rd set of benchmarks had an overclocked video card. Can you your scores at stock GTX speeds for better cpu comparison?

Unfortunately I'm currently using single-core (Opteron 146) so the comparison wouldn't be any better. I ran the 3DMark with a loaner CPU. (Thanks nealh)

 

[TuxDave]

Originally posted by: lopri
Hate to quote HardOCP but they have now a front-page banner that says things in line with my observation.

Originally posted by: HardOCP
Real world gameplay analysis of Core 2 vs. Athlon 64 coming very soon. Shows different results than canned benchmarks? You betcha

Be sure to let us know when the article is released.

 

[Griswold]

Originally posted by: theprodigalrebel
Comparing the CPU scores for 3DMARK,
Conroe @ 2.93GHz: 2517 Marks
X2 @ 2.8GHz: 2039 Marks

~23% increase for the Conroe. Isn't this kinda good? Or does this test, like all synthetic tests, reveal nothing concrete about real-world performance?

And shrinks to ~12% when running the X2 at the approx. same clock speed.

But yea, synthetic is synthetic. :thumbsdown:

 

[JEDIYoda]

Originally posted by: broly8877
Faulty temp sensors, guys.

Almost all boards are showing 10-25c more than what they actually are.

Remember, Intel says the max the chip can handle is ~65c, so that reading is obviously not right.

Faulty temp sensor or some amd fan boy spreading alot of dribble...

 

[Hard Ball]

Originally posted by: JEDIYoda
amd fan boy spreading alot of dribble...

How do you figure that? You are saying that he faked the screen shot??

If that's what you think; how do you know that any of the previous benchmarks from any of the Conroe ES would not be a fake?

 

[RichUK]

Well im fully satisfied with my rig (thnx Lopri for the GFX card and CPU ), i have no interest in Conroe. Looking forward to people's OCing results though.

 

[Kyanzes]

Originally posted by: lopri
the results he got seem to be a somewhat mixed bag, as far as 3D gaming is concerned.

How could you say anything remotely like that?

Let's check it out:

Awesome Conroe:
CPU2 - Red Valley 1.275 FPS

The outdated X2:
CPU2 - Red Valley 1.023 FPS


I mean, man, you really should reconsider! Not convinced?

Check this out also:

Awesome Conroe:
3DMark Score 10230 3DMarks
The obsolote X2:
3DMark Score 9252 3DMarks

It's a breakthrough man... Seriously though, it's not about the CPUs really, as you said, but the vidcards. There are no miracles.
But we have to wait until extensive, reliable and - most importantly - widely confirmed tests start to appear on the review sites.

 

[lopri]

Originally posted by: TuxDave
Be sure to let us know when the article is released.

Here

 

[imported_Crusader]

Waiting for reverse hyperthreading, if not on 939.. then whoever has the better price between AM2/Conroe (with RHT) and best availability when RHT is released for one or both of the two.

A64@2.5ghz is plenty for me and for my GPU.

 

[terentenet]

Originally posted by: coldpower27
Oh please, comparing a overclocked GPU vs a stock 1.... 3D Mark 2006 is GPU bound.... 1000 Point with the same GPU subsystem is failry impressive.

No, 2006 is limited all over. I have a 7900GTX SLI setup running stock speeds, and the SLI load balancind shows that the video cards RARELY hit 80% usage. That only means that the CPU (X2 4800) can't keep up to feed them. I bet a better CPU really makes a difference in 3dmark2006. I get 1823 CPU score, 8620 3D marks.
Conroe gets 2500 CPU score, so it's about 40% more efficient. RESPECT, but I'll stay with AMD. C'n'Q enabled, 32 idle, 48 full load.
I still wonder when we'll get to see a processor that now runs the CPU tests at 0-1 FPS with 20FPS. in 2010 maybe? 6GHz 16 cores anybody?