8800GTX Benchmarking Results

I would like to give everyone considering the new 8800GTX a nice benchmark session to display how it scales with CPU processor raw clockspeed.

System Specs:
E6700 Conroe "week 27"
OCZ Titanium Alpha VX2 2-Gigs
Silverstone Zeus ST56ZF Powersupply
EVGA 8800GTX

First lets take a look at the EVGA packaging:

http://svtsnake.com/ocforums/box.JPG

http://svtsnake.com/ocforums/box2.JPG


And here are some snapshots of the GPU in use:

http://svtsnake.com/ocforums/card1.JPG

http://svtsnake.com/ocforums/card2.JPG

http://svtsnake.com/ocforums/card3.JPG

http://svtsnake.com/ocforums/card4.JPG

The idea is to test out a brand new EVGA 8800GTX gpu at stock core/mem speeds. Only to adjust the clockspeed of your actual CPU on the motherboard. I tested an E6700 Conroe processor all the from 2.4 to 4.2 GHz at 4 different measuring points. Here are some of the results:

E6700 Conroe @ 2.4 GHz
3DMark05

http://svtsnake.com/ocforums/24003dm05.JPG

3DMark06

http://svtsnake.com/ocforums/24003dmark06.JPG

E6700 Conroe @ 2.93 GHz
3DMark05

http://svtsnake.com/ocforums/29303dm05.JPG

3DMark06

http://svtsnake.com/ocforums/29303dm06.JPG



E6700 Conroe @ 3.3 GHz
3DMark05

http://svtsnake.com/ocforums/33003dmark05.JPG

3DMark06

http://svtsnake.com/ocforums/33003dmark06.JPG




E6700 Conroe @ 4.2 GHz
3DMark05

http://svtsnake.com/ocforums/42003dmark05.jpg

3DMark06

http://svtsnake.com/ocforums/42003dmark06.jpg


And here is the final analysis on how the GPU scales when raising CPU clockspeed:

http://svtsnake.com/ocforums/3dmark2005chart.JPG


http://svtsnake.com/ocforums/3dmark2006chart.JPG


After actually reviewing and producing these results I have now fully turned my opinion and position around on this matter. I am in agreement that the CPU is not a big GPU bottlneck as I previously stated. Thats what the whole idea of testing is about. I jumped the gun based on 3dmark05 results and myself being overly happy with the card but now, testing proves otherwise. If you take a close look at the SM3.0 difference in performance, it clearly does not scale with CPU power well at all. From 2.4 GHz to 4.2GHz you are looking at a 100 point increase if that. Myth busted = 8800GTX is not as CPU bound as we make it out to be.

***OFF TOPIC UPDATES***
First attempt at Overclocking it. I tried setting core over 620 and nv control panel said it failed core test. So, I am taking it slow.

Maxed out the fan to 100%
upper core to 614
upped mem to 920 (1840mhz)

CLICK FOR IMAGE: http://svtsnake.com/ocforums/OC1.JPG

Broke 20,000 3dmarks!!

Open Ntune
Set Fanspeed - 100%
Up core to 625
Up mem to 950 (1900mhz)

http://svtsnake.com/ocforums/OC2.JPG

FEAR RESULTS, OVERCLOCKED AT 625 core, 950 mem. Please note, FEAR IS ON ULTRA/MAXIMUM SETTINGS and every single graphical option including AA and AF are set to maximum.
Minimum FPS = 77 LOL
Average FPS = 159
Maxmimum FPS = 440

This card has ungodly power.

CLICK HERE FOR SCREENSHOT: http://svtsnake.com/ocforums/FEAR2.JPG


OMG....
The High Quality Results are just stupid fast. HQ is 1 step down from Ultra/Maximum quality.

Minimum = 117 FPS
Average = 256 FPS
Maximum = 679 FPS

JUST ABSOLUTELY STUPID FAST

CLICK HERE FOR IMAGE OF SCREENSHOT: http://svtsnake.com/ocforums/FEAR3.JPG


UPDATE with some more good results.
The card seems to be overclocking very nice at stock voltage an stock air cooling.

Opened up NTune
Set fan to 100%
Set core to 640 mhz
Set mem to 1000 (2000mhz)

CLICK HERE FOR SCREENSHOT: http://svtsnake.com/ocforums/OC3.JPG

This EVGA card overclocks incredibly!!
But I am now announcing my first screen freeze and WIndows XP reboot.
Opened Ntune
Set fan to 100%
Set core to 650mhz
Set mem to 1100mhz (2200mhz effective)

I think I finally asked too much of this DDR3. I am going to run my next benchy at mem 1050 and core 650... see if it was definitely the memory holding me back. But it got through half of return to proxycon and then failed.

Give me another 10 minutes or so.

Yup, I was right. Asking DDR3 of 2200 effective mhz was too much to ask of this card, stock cooling, stock volts. 650 core worked great.

Updated Overclock:
Open Ntune
Set fanspeed to 100%
Set core 650 mhz
Set mem 1040 mhz (2080 effective mhz)

CLICK HERE FOR SCREENSHOT http://svtsnake.com/ocforums/OC4.JPG

So far, my best stable OC seems to be right around 650 core and 1040 mem. I will play around with core alone, while leaving mem at stock to see how high the chip can actually go.

Here is a 3dmark06 shot with 650 core and 1040 mem:

CLICK HERE FOR SCREENSHOT: http://svtsnake.com/ocforums/OC5.JPG

Lates Update:
Bumped up clocks:
Mem 1040
Core 652

CPU Clockpseed - 4.25 GHz

I am really pushing for that 21,000 mark now!

CHECK OUT SCREENSHOT HERE http://svtsnake.com/ocforums/OC6.JPG

21,000 IT IS!!!!
I went all out for this one. Bumped up CPU clocks to 4.35GHz
Open Ntune
Fan to 100%
Set core 653 mhz
Set mem 1040 (2080 effective)

http://svtsnake.com/ocforums/OC7.JPG

Very nice with a cpu scaling test, most helpfull and interesting. Nice work mate :)

very nice thread sir, thank you for your hard work.. or play :D info appreciated

Good tests :toast:

I dont know how you can conclude that the G80 is not CPU limited however.....

The PS3.0 test is simply theoretical and the output of the pixel shaders probably always run at peak "potential" capacity.

When the rubber hits the road however, you need the extra cpu cycles to effectively transfer this pixel processing potential into usable rendered frames.

To me, in 3dm05, firefly forest is the most demanding test. Seeing the frame increases in this test alone as cpu frequency scales is impressive.

I am now realizing that a projected 3.6ghz oc on a cornrow may not be enough to fully juice the g80.

Also-- when and where did you buy the evga? I have one inbound from the Egg and was hoping to get the evga black sticker (like on the box) on the front of the heatsink instead of the reference green sticker.

very nice post and helpful information. good work.

Quote:

---

Originally Posted by dominick32

After actually reviewing and producing these results I have now fully turned my opinion and position around on this matter. I am in agreement that the CPU is not a big GPU bottlneck as I previously stated. Thats what the whole idea of testing is about.

---

Most of the big CPU bumps I see is going quad core from dual core. Its really apparent if your doing SLI on the 8800's. For some reason though 05/06 (06 has a cpu test but not 05) seem to do better with more CPU while 03 doesn't. You get a pretty big 03 boost from SLI without going crazy on the CPU.

Eric

nice testing, great work. I should be doing some similiar benchmarking this coming weekend, but with my e6600/ week27. :)

Not cpu limited ?


Each time you OC the CPU, each time the score rise, so ? ... ;)

Yes by not a 50% improvement, but a good few points at each time :D

Sli of 8800GTX need more than a simple E6600 even X6800 :slapass:

If the aim is to explode some scores at 3DM06 :banana:

Very nice, gives me a nice idea on what I should aim for.

Well done, dominick32

Thanks Guys!

I have an update. Broke 21,100:

Set fan 100%
Set core 660
Set mem 1039 ** This seems to be my maximum stable GPU core/mem overclock** So I decided to raise the E6700 conroe to 4.37GHz:

Nice progression there....it really helps me to see what my 8800's will do once I have then installed. I am an AMD owner (nice FX-60. . .capable oc'er) but I will also try to see what I can do with just one card and compare my progression to the SLI results when I add the second card. Either way, gaming @ 1920x1200 with AA/AF maxed out and all eye candy turned up will be sweeter than sweet!

Nice Job M8 :)
Keep up the good OC :P

First Time Poster!

Anyway your score for 3dmark05 with your e6700 Underclocked to e6600 speeds shows that the e6600 is a bottle Neck to a 8800GTX. The reason why Im saying this is because I scored 14044 with a EVGA 8800GTS at stock with a e6600 and 2GB of corsair Dominator pc6400-C4D. http://service.futuremark.com/orb/re...=0&UID=7126333

Quote:

---

Originally Posted by dominick32

Thanks Guys!

I have an update. Broke 21,100:

Set fan 100%
Set core 660
Set mem 1039 ** This seems to be my maximum stable GPU core/mem overclock** So I decided to raise the E6700 conroe to 4.37GHz:

---

Rise pci-e bus to over 110mhz and set little lod, with 350mhz more, than my Air cooled E6600, you should get 22k 2005;)
My best run was with older version of 3dmark :(
21238
http://img148.imageshack.us/img148/6...1070ho9.th.jpg

And now lower validation with the new patch
21165
http://service.futuremark.com/compare?3dm05=2408972

Not CPU bound?

How do you explain this?:

http://svtsnake.com/ocforums/3dmark2005chart.JPG

Going from 2.4GHz to 4.2GHz almost nets you a whopping 5K.

Nothing to sneeze at in my book.

A 5GHz Conroe still bottlenecks a highly overclocked 8800GTX SLI setup, look at K|ngp|n's 2001 nature tests.

Bumping the CPU to 5,2GHz gained a 100fps in nature!

How's that for a bottleneck? :D

Nifty stuff.

yeah i think your conclusion is 50% right, pretty nice overclocking log i must say, i am hoping to have the time to do that kind of report with my setup too. To say that a cpu is not bottlenecking, you would have to see minimal performance boost between 2.4ghz and 4.2ghz lets say... but youre benchmark really show a huge performance leap.

i just sold my 4400+ oc @ 2.6ghz and 2gb of pc4000 and i advised the buyer not to by 8800 series because the pc wont be able to push enough processing power to follow the 8800. im pretty sure a 8800gtx over a 7900gtx on that setup will not show any diff

Hi,
There are a lot of reviews/benchmarks aboutthis new stuff from nV.
All of them are talking about few things: Shaders, Pixels...and all of benching are done on the games.
Now, Im asking myself, what are improvements in 3dsMAX?
Better image quality?
Better AA?
D3D or MAXtreme D3D 9?
So, here is old benchmark from even 3dsMAX 4.0, Im shure you remember it.
If someone already own some 8800 card please do some benches!!!
Here's the link for downlaod, rar archive includes all needed scenes+script required for running bench.

http://rapidshare.com/files/3108339/bmark4.rar.html

I didnt use it on MAX9, so I don know how it works, bit on MAX8 it works 100% (sometimes you have to run it twice time to run it reguraly).
Also, all of the benches should be done in 1280x960, just because of comparing w/old graphic card benchmarkings.
If you find it interesting you can also done it in 1680x1050 as we all easly move to this "industry standard" wide screen.

THANX!

I think we all would have benefiet from this.
Just to say, I do not believe to SPEC benches, they are sinthetic, even SPECfor MAx is not realworld bench, here you have pure poligons&textures.


PS-If you find message "Tetures did not find etc..." just click CONTINUE, or find your old MAX 5.0 install CD!

3DMARK06
I've had the best score running the 8800GTX @ 680/2020, FX62 3190MHz.

11120
http://service.futuremark.com/compare?3dm06=853193

With your CPU you should get into the 15,000 or higher

I get 14.800 with e6600@ 4,0 and hardmod xfx8800gtx 700/2200 1,55vcore (vr-zone thanks 400ohm soldier;)
nice 3400cpu points in 06mark

Quote:

---

Originally Posted by Janbatos

I get 14.800 with e6600@ 4,0 and hardmod xfx8800gtx 700/2200 1,55vcore (vr-zone thanks 400ohm soldier;)
nice 3400cpu points in 06mark

---

Link PLS !
13.8 sounds more fair, but doubt,they also can be reached with so low clocks

With 3400 CPU points and 675/2160 mhz my tweaked result is little over 13k
http://img148.imageshack.us/img148/3...2140as7.th.jpg

let's do the Math:))
3400cpu
to get final 14800
you need
SM2 6800
SM3 6850
;)
No way, man , never @700/2200

sorry 750/2200 waterchilled gpu tested;)

Quote:

---

Originally Posted by Janbatos

sorry 750/2200 waterchilled gpu tested;)

---

I said it, but will repeat it . .. LINK PLS.
Even @750/2200 SM2 and SM3 over 6800 points are not possible.;)

21k 3D05, thats about what I get in 3D03 :D