AMD Shanghai/Deneb Review Thread

[Zucker2k]

Well this is what i get if i set affinity to ONE CORE, with view distance 22 and detail distance 70 @ 3ghz. So if i7 has 8 threads to toy with....not that difficult to see how you could get 22% with 60fps when there is the core shutdown stuff. But regardless it's not about how much of the CPU is utilized it's the performance you get out of it.

8 threads is not the same as 8 cores; besides the game has to be able to utilize 8 threads. Let's assume it does, that's still 44% compared to 70%. The point I'm trying to make is an efficiency, clock/clock in game comparison. The available headroom on i7 is huge and should not be ignored since this is a comparison.

[demonkevy666]

8 threads is not the same as 8 cores; besides the game has to be able to utilize 8 threads. Let's assume it does, that's still 44% compared to 70%. The point I'm trying to make is an efficiency, clock/clock in game comparison. The available headroom on i7 is huge and should not be ignored since this is a comparison.

maybe it's the fact that is reads 4 core so it thinks well here 4 threads. so since it's has 8 threads but only 4 cores. maybe the code isn't totally working out there is 8 threads, because it's only 4 cores.

need to see this game a 8 core server system with Shanghai's too then.

[ajaidev]

8 threads is not the same as 8 cores; besides the game has to be able to utilize 8 threads. Let's assume it does, that's still 44% compared to 70%. The point I'm trying to make is an efficiency, clock/clock in game comparison. The available headroom on i7 is huge and should not be ignored since this is a comparison.

You are quite wrong the game "GTA4" takes HT as added cores as in total of 8 CORE's on an i7 The reason why it does not show up as 42 instead of the seen 21 is that HT just sits on its arse doing noting "HT will do noting great if enabled or disabled"

So the game thinks its running on a 8 Core computer and compute scores for all 8 so called cores or threads. So if first on is at 40, second at 40, third at 40 and forth also at 40 but the HT all at 0, this will give you a result of amm 20% overall utilization.

Now besides this you do know that the game IDs my Q6600 as a PIII Xeon

[Zucker2k]

You are quite wrong the game "GTA4" takes HT as added cores as in total of 8 CORE's on an i7 The reason why it does not show up as 42 instead of the seen 21 is that HT just sits on its arse doing noting "HT will do noting great if enabled or disabled"

So the game thinks its running on a 8 Core computer and compute scores for all 8 so called cores or threads. So if first on is at 40, second at 40, third at 40 and forth also at 40 but the HT all at 0, this will give you a result of amm 20% overall utilization.

Now besides this you do know that the game IDs my Q6600 as a PIII Xeon

Are we assuming that they run the benchmark with HT on then? Because that'll usually lead to lower scores/fps.

[Lightman]

Thanks iocedmyself, great work you're doing here. That first score at 3.6Ghz didn't look pretty for the phenom II though, as it was 22% percent i7 (no one in their right mind would turn HT on for gaming, that anomally is well established), vs Phenom II 70%. Am I the only one noticing that huge gap? Are you sure no other process was stealing huge cpu cycles? Good job man, keep up the good work.

Simple explanation:

GTA IV can spawn 3 threads which can utilize CPU fairly well. For that reason Single, Dual and Triple-Core will show CPU utilization close to 100%!
Quad-Core will be utilized in 75% assuming no other thread is using it and GTA IV will extract everything from remaining 3 cores.
Six-Core CPU will hit max. 50% utilization.
8 cores/threads can be utilized 37% by GTA IV!
Now Core i7 is not hitting close to theoretical 37% and the reason for that can be either HT lowering utilization or GPU saturation (in other words CPU too fast for a given GPU).

Is that clear enough?

[gamervivek]

3.0 ghz, 2ghz NB 2ghz HT 47 draw distance 100 detail distance 100 car density 16 shadow density 1920x1200 = 40.19fps

3.6 ghz, 2ghz NB 2ghz HT 34 draw distance 100 detail distance 100 car density 16 shadow density 1920x1200 = 31.97fps

did you turn of a core or two at 3.6?

[gamervivek]

the draw distance is most important for cpu.
very HARD setting !

as i see it,lesser the draw distance lesser the fps.either this benchmark is BS or there was some other factor reducing the framerates.

[Zucker2k]

Simple explanation:

GTA IV can spawn 3 threads which can utilize CPU fairly well. For that reason Single, Dual and Triple-Core will show CPU utilization close to 100%!
Quad-Core will be utilized in 75% assuming no other thread is using it and GTA IV will extract everything from remaining 3 cores.
Six-Core CPU will hit max. 50% utilization.
8 cores/threads can be utilized 37% by GTA IV!
Now Core i7 is not hitting close to theoretical 37% and the reason for that can be either HT lowering utilization or GPU saturation (in other words CPU too fast for a given GPU).

Is that clear enough?

Are you sure those settings are maxing the GPU? I don't think those settings are very gpu-bound. In any case, your last statement takes nothing from what I said; if the i7's HT is the culprit, meaning the cpu is taking a negative hit, the results are still better than the PHII And if the gpu is maxed, then are you saying it takes only 22% of the processing power of a Ci7 at 3.6ghz to max the a 4870x2 with GTA IV? Or are you leaning towards the unknown?

[noki]

The cpu score in 3dmark 06 is very low. I saw yesterday an screenshot of 3dmark 06, with a phenom II 940 @ 3,8Ghz, 5700 pts in cpu score, here:

http://www.xtremesystems.org/Forums/...=210814&page=2

Then the score of iocedmyself with his phenom II 940 @ 3,6Ghz is the same that you would get with a phenom 9850 @ 3,3 GHZ, here:

http://service.futuremark.com/result...eResultType=14

So I wonder what's wrong, iocedmyself's ram timings are not good but I think the score should be better than that, what do you think guys?

[iocedmyself]

I mistakenly thought i had shadow density and car density maxed in the 3.0 one, i did not, they were maxed in the 3.6 which creates a heavy performance hit.

as for 3dmark 06 i was benching at 1920x1200, not the default 1280x1024, will rerun default momentarily

[crazydiamond]

on vista 64 (not tuned to performance at all) i get ~4500 cpu points on 3dmark06 with cpu@3.36ghz nb@2400 mem@950/4.4.4.12 and bumping cpu up to 3.44ghz in only get ~4600 cpu points... i was getting ~5000 on xp32 @ 3.36

[ajaidev]

Are you sure those settings are maxing the GPU? I don't think those settings are very gpu-bound. In any case, your last statement takes nothing from what I said; if the i7's HT is the culprit, meaning the cpu is taking a negative hit, the results are still better than the PHII And if the gpu is maxed, then are you saying it takes only 22% of the processing power of a Ci7 at 3.6ghz to max the a 4870x2 with GTA IV? Or are you leaning towards the unknown?

Well off course i7 is faster than a PhII no one expected anything else "Clock to clock" but as OP has stated the PhII has a OC advantage over the i7

The HT does slow down some game and speeds up others while on GTA4 i dont think it has any type of impact. As i stated above i think even after disabling HT GTA4 will read the CPU with 8 imaginary cores!!!

22% means real 44% cpu utilization see my post above.....
Still in price to performance Phenom II is better as compared to i7 platform

[Oliverda]

iocedmyself

What are the BIOS default clocks for NB and HT?

[Zucker2k]

Well off course i7 is faster than a PhII no one expected anything else "Clock to clock" but as OP has stated the PhII has a OC advantage over the i7

The HT does slow down some game and speeds up others while on GTA4 i dont think it has any type of impact. As i stated above i think even after disabling HT GTA4 will read the CPU with 8 imaginary cores!!!

22% means real 44% cpu utilization see my post above.....
Still in price to performance Phenom II is better as compared to i7 platform

If HT is turned off, Windows can only 'see' 4 cores (or threads). Do you know something we don't? If GTA IV can only run on 3 threads, what's the use to enable HT? To cheat?

[ajaidev]

If HT is turned off, Windows can only 'see' 4 cores (or threads). Do you know something we don't? If GTA IV can only run on 3 threads, what's the use to enable HT? To cheat?

http://www.xtremesystems.org/forums/...&postcount=385

[Zucker2k]

http://www.xtremesystems.org/forums/...&postcount=385

What's your point? That's the Phenom II benchmark, and the chip has no HT (read hyperthreading).

[ajaidev]

My point was that setting affinity to ONE CORE results in similar results as i7.

Now, that means per core i7 gets 11% utilized as compared to 19% for the PhII.

[Zucker2k]

My point was that setting affinity to ONE CORE results in similar results as i7.

Now, that means per core i7 gets 11% utilized as compared to 19% for the PhII.

Okay, I'm not going to drag this on any further since you can't see my point. What's the use of crippling the cpu, then try to validate the result?

[Lastviking]

Okay, I'm not going to drag this on any further since you can't see my point. What's the use of crippling the cpu, then try to validate the result?

Its more that you cant understand his point or are not willing to see it. Its his point and not yours again...

Plz stop being rude to another users , Go back to the intel forum plz!

[iocedmyself]

Well off course i7 is faster than a PhII no one expected anything else "Clock to clock" but as OP has stated the PhII has a OC advantage over the i7

The HT does slow down some game and speeds up others while on GTA4 i dont think it has any type of impact. As i stated above i think even after disabling HT GTA4 will read the CPU with 8 imaginary cores!!!

22% means real 44% cpu utilization see my post above.....
Still in price to performance Phenom II is better as compared to i7 platform

Yes i have to agree that the HT is most likely to be the culprit for 22% cpu load, either that or the power saving feature is disabling several of the cores and still counting it as 4 physical or 8 logical threads. Easy way to tell, look in task manager in the performance tab, if it is showing 8 devisions than GTA is calculating based on 8 cores.

I7 has better ram, but gimped caches horrible power consumption (total system draw, not counting the disabled core crap) It's expensive for little real world performance gain and there is little if any overclocking room.

We're looking at raw performance here in gaming not efficiency. There is at 3.6ghz comparison results showing i7 has less than 10% performance gain over phenom II regardless of cpu load. The difference is that 3.6 is the glass cieling, or damn near for i7 where AMD still has anywhere from 400-1000mhz headroom to overclock....with air cooling alone. Plus anyone that is going to be using an I7 or a phenom II at an early stage, especially with a 4870x2 is going to have considerably better than a 1280x1024 display i would think.

Upgrading to I7 means new cpu, heatsink, motherboard as well as DDR3.

Phenom II means....if you already have an AM2+ board you buy a phenom II and at most update your bios. If i had to spend the cash to upgrade.....i know what would be more appealing from all angles

Luckily i won't have to trouble my mind with such thoughts

But Zucker i do see your point and whether HT is enabled or not on the I7 is just guess work on ALL our parts. As well as the fact that things like car density and shadow density settings are unknown and those could be set at 0 which would easily lighten CPU load. But then again the system specs are completly unkown based off that screeny so not really much is certain, if i missed a further detailed explination of the system that bench was done on please correct me, not trying to bait anyone or deny any evidence, just admiting that i've been somewhat preoccupied and using only that screeny as a basis there is a lot left up in the air that could effect the results.

As to the point of enabling HT and then "crippling the cpu" With the way the I7 power saving features work it could just disable some cores, GTA takes into account the existence of all of them and performance results are....creative. They could be accurate but i don't think anyone can deny the fact that when it comes to intel benches, they have not always been on the level in the past. Nor the fact that 22% cpu load at 60 FPS is going to impress people, and if they can show those results why wouldn't they? Besides it may TECHNICALLY be true, but it may be taken out of context as to how those results were obtained

But i can't deny that intel may have improved ability to offload more to the GPU at lower resolutions which also could account for the low utilization. If you'll notice Cpu load on phenom II drops from 70% at 12x10 to 57% at 19x12 at same detail settings, and higher resoltions are more what im interested in.

At any rate not trying to get into a flame war, just trying to shed some light on Denebs as very best i can independent of I7 numbers, and speculation will almost certainly result in hair pulling and name calling with no one winning.

Oliverda, Bios default clocks for NB is 2ghz, HT is also 2ghz @ 200x10

My heatsink through out all testing has been TT DuoOrb http://www.thermaltakeusa.com/Produc...1&ID=1748#Tab4

Though i've finally decided that it is time to break down and break into the realm of water cooling, So i'm about to invest in this Swiftech Apogee GTZ water block

http://www.martinsliquidlab.com/Swif...TZ_Review.html

http://xcpus.com/GetDoc.aspx?doc=69&page=1

And the Swiftech MCP655 adjustable Pump
http://www.bjorn3d.com/read.php?cID=1363

So I should hopefully get around to putting this chip of many shines under water definitely before xmas, hopefully ill have it running by the coming weekend so i can shoot for a stable 5ghz.

After new years, i'll have a more fun endeavor to undertake...well two actually. The more certain of tasks will be setting up a water chiller for the Reservoir with a pair of under-volted 12730 each will have power req at around 8V drawing 12A and moves about 40W across 30C. heat load is about 160W. with a 5 pound nickle plated copper Melcor sink that is a rise of 9C. Two should do a fine job on a deneb... around a 300W heat load total, control for 5C water.

The less certain and more fun task will be the old joy of slapping a TEC right on the CPU (btw 3.0ghz at 1.1v has a TDP of around 55-60w) So that water block over a beefy TEC will yield some negative C

[Oliverda]

Oliverda, Bios default clocks for NB is 2ghz, HT is also 2ghz @ 200x10

Thanks, and the default Vcore is 1.35V if I'm not mistaken. Am I right?

[Adonay]

I wouldent take those % cpu useage for granted as GTA 4 is a pretty buggy game, the last bechmark i would trust would be a GTA4 other than the fps ingame useing Fraps and some kind of cpu monitor.Have anyone of you noticed it only shows 3 cores in use within windows btw?.. Anyways looks like i will go for amd next as a spring project ... I wonder how well the ddr3 wil inpact the performance as well ...

[Zucker2k]

@iocedmyself; thanks for everything. I just saw something I thought I might bring up. I just wish some on these forums have the objectivity you've just demonstrated. I'm not here to bash any cpu, or manufacturer, just to point out the facts as I see it. Kudos fr your efforts; this is how peeps make names for themselves, by showing us the early figures, though admittedly, even these would improve over time as process matures.

[Caveman787]

If memory performance is good ddr3 will allow massive bandwidth along the lines of tri channel ddr3 for nehalem.

Yes the default vcore from all those with samples has been shown to be 1.35v...some of you need to read a few pages back before you ask questions.

Not trying to be mean, but some ?'s have been answered time and time again through the thread both in text and screenshots.

[LC_Nab]

Its sometimes pointless to compare amd and intel , it always ends up into a AMD vs Intel , which we are just trying to see performance between the 2 chips many would like to know how phenom II performs , so please ppl take the Vs to another thread and leave this one alone for ppl who wants information about the new Phenom II . Personally I just want to know its + and - . and thank you all for the info so far . So please intel lovers " no pun intended " let the AMD gurus do their job so we can enjoy the new phenom II . And caveman , let the guy ask questions theres nothing wrong with it , not everyone has the time to scroll 17 pages of info .

Again good job guys on the phenom II info keep up the good work .