HD5870 vs GTX295

[Setix]

Hey guys,

I'm building a new rig and have come down to shopping for graphics cards, so heres my question.

What would be the better buy in terms of performance, reliability and support?

2 x hd5870's in crossfire or 2 x gtx 295's

I'm not looking for fanboi replies or biased responses. I am really interested in an actual logical, well reasoned argument debating one or the other. These cards will be OCed, liquid cooled on a x58 Classified motherboard.

I do know that the "end of life" order has gone out on the Gt200b chips. So I am concerned about getting a replacement if a card blows out. But havent heard much about the new Fermi chips in regard to gaming performance.

Any help?

[Splave]

if you can find 1 5870 youd be lucky lol

[Setix]

if you can find 1 5870 youd be lucky lol

Newegg is expecting a shippment at the end of this week, so who knows...but if NVidia has a better card releasing early next year I might just wait

[TheGreenFoX]

I would take two 5870.
Reason 1. Fewer GPUs gives better overall performance.
Reason 2. DX11
Reason 3. Fewer watts

[zalbard]

Wait I guess.
295 isn't the best buy in any case since it only supports DX10, is a dual GPU, so the more you add the worse it scales, and more importantly, it is already VRAM volume limited in a few cases.

[billdavis]

if you can find 1 5870 youd be lucky lol

i just bought 2 more yester day. you need to camp newegg they come up randomly in groups of 5 or so

[Piotrsama]

The 5870s are very good OCers.

[Splave]

i just bought 2 more yester day. you need to camp newegg they come up randomly in groups of 5 or so

spam my pm box next time you see some rofl

[Setix]

I suppose I'll wait until Fermi and Hemlock are released and we actually see some benchmarks. Its not critical at this juncture, I have other parts I can buy in the mean time.

how many games are actually using Physx?

[MetalWaterBox]

I would take two 5870.
Reason 1. Fewer GPUs gives better overall performance.
Reason 2. DX11
Reason 3. Fewer watts

PLUS they are cheaper:up:

[zalbard]

I suppose I'll wait until Fermi and Hemlock are released and we actually see some benchmarks. Its not critical at this juncture, I have other parts I can buy in the mean time.

how many games are actually using Physx?
PhysX is unimportant, not many games use it, and even less are going to do so since DX11 Compute Shaders and OpenCL implementation is going to work for both vendors, and not just one.

[Chickenfeed]

Depending on the games played and whether you are into F@H or not, I'd say for single card, a 295. However given the investment involved with 2 cards at this point, 2 5870s hands down. You'll have roughly half the power consumption not to mention much less noise as well as better scaling ( 2 5870s will scale more consistently and have less issues than a 3-4 gpu setup ) Don't forget to take into account the resolution you plan to play at as well. 2 5870s should spank any upcoming games (as well as current games) at 1920x1200 with 8x AA where the same setup should be fine with 2560x1600 with no to 4xAA well into the future ( assuming VRAM doesn't become a limitation )

The only real incentive I personally have in choosing an Nvidia gpu right now are their drivers and overall game performance. They still put AMD to shame in this regard ( I'm talking their control panel, profiles as well as multigpu scaling across more titles )

[texasreefer]

I agree.....ATI is awesome, but nVidia overall game performance has always had less hitches. On the other hand, my bud just got his second 5870, and it perform almost equal to my 4870X2s in crossfire e.g. I hit 30k in 3dmark 06 and he hits 29.4k, I hit 24.7k in Vantage he hits 23.7k. That's just a generalization, but he has an I7 @ 4.2 and I'm at 4.4.

I wouldn't necessarily go with a gtx295 at the moment, considering the next generation nVidia chips are coming out (and I'm kind of excited to see what they're going to put out). nVidia always seems to match or exceed what ATI puts out. If you get a 5870 now you would have no problem getting rid of it for as much as you paid or just a bit less, can't say the same of a 295.

[Carfax]

4.4ghz and you're on air? What voltage are you using, and what are your temps?

[Jor-El]

PLUS they are cheaper:up:

Exactly. The performance difference isn't even close to being comparable to the price difference. This is a no-brainer if you ask me except to the most die-hard fans of the green guys.

[dan7777]

I agree.....ATI is awesome, but nVidia overall game performance has always had less hitches. On the other hand, my bud just got his second 5870, and it perform almost equal to my 4870X2s in crossfire e.g. I hit 30k in 3dmark 06 and he hits 29.4k, I hit 24.7k in Vantage he hits 23.7k. That's just a generalization, but he has an I7 @ 4.2 and I'm at 4.4.

I wouldn't necessarily go with a gtx295 at the moment, considering the next generation nVidia chips are coming out (and I'm kind of excited to see what they're going to put out). nVidia always seems to match or exceed what ATI puts out. If you get a 5870 now you would have no problem getting rid of it for as much as you paid or just a bit less, can't say the same of a 295. 24k with crossfire in vantage seems low i get 26k with a 295 physx on with it off i get 22k i7 @ 4.2

[Jamesrt2004]

24k with crossfire in vantage seems low i get 26k with a 295 physx on with it off i get 22k i7 @ 4.2

exactly somethings off, 5870 normally withn 10% of 295, not 2 5870 loosing to it.. lol somethings off, but oh well

ims ure drivers and all that poo will be fixed soon :up: tho im still waiting for nvidia, to have a look and just for pricedrops really

[spajdr]

I would take two 5870.
Reason 1. Fewer GPUs gives better overall performance.
Reason 2. DX11
Reason 3. Fewer watts

2. but what games? dirt 2 is coming soon and thats pretty much all, before it get widely used, nvidia will have his card released too and who knows how it will perform, maybe better
Heaven benchmark shows nice performance hit when tesselation is in action.

[texasreefer]

Carfax......I have a DO stepping I7 my voltage is at 1.365, my idle temp is 32C and load 58C, my QPI is at 210 and Mult at 21......believe it TRUE Copper baby. Oh yeah, and my X2s idle at 40C.


Dan7777 ..... Thanks, that has to be it then. You get 4k just by enabling physX, Vantage does not scale right with ATI. These are my scores. Those are the scores that my buddy with the 5870s is posting, he says he's overclocking the shissnit out of those cards. The thing that threw me off the most was that he was hitting 26+k with one card, then when he added the second only got a little over 3k boost. I've been using and building since the damn AMD anthalon days and when FX processors were the shoot, I don't think I'm missing anything.
I seemed like I got more of a boost than that when I was using an HD3870 and went into crossfire.

The 295 is still the top benchmarker, and I have no reason to make up numbers as I am not trying to sell anything. Right now the only thing I can think of to fix it is to figure out how to run physx....I've been looking for a cheap 9800gt to run CUDA. Only problem is DAMN....these things Still aren't cheap. I can't find one for less than 99 USD...that says a lot for nVidia. If you guys can show me a different way to set up Vantage...or maybe a patch that I haven't found, I'd definitely be allll ears to that. I have win 7 in the mail, and I'm looking for a used 9800gt on Ebay. I just reaally hate having to do all that just to boost up a benchmark.

[Carfax]

Carfax......I have a DO stepping I7 my voltage is at 1.365, my idle temp is 32C and load 58C, my QPI is at 210 and Mult at 21......believe it TRUE Copper baby. Oh yeah, and my X2s idle at 40C.

No offense dude, but I have to see some evidence to believe that. It's not the clockspeed that makes me raise an eye brow, it's the temps.

The True Copper is a great heatsink, but it is equaled and even outperformed in many cases by the Prolimatech Megahalems which I have in my system, and my load is around 75C at 4ghz @ 1.28v.

Most air cooled rigs I've seen have similar temps at that clockspeed.......but nothing like what you're saying. Those temps seem more likely on a high end water cooling set up.

Although admittedly it could also be that you keep your computer in a very cold environment..

[Jamesrt2004]

No offense dude, but I have to see some evidence to believe that. It's not the clockspeed that makes me raise an eye brow, it's the temps.

The True Copper is a great heatsink, but it is equaled and even outperformed in many cases by the Prolimatech Megahalems which I have in my system, and my load is around 75C at 4ghz @ 1.28v.

Most air cooled rigs I've seen have similar temps at that clockspeed.......but nothing like what you're saying. Those temps seem more likely on a high end water cooling set up.

Although admittedly it could also be that you keep your computer in a very cold environment..


tbh its better then most high end watercooling setups on here.. lol

[One_Hertz]

No offense dude, but I have to see some evidence to believe that. It's not the clockspeed that makes me raise an eye brow, it's the temps.

The True Copper is a great heatsink, but it is equaled and even outperformed in many cases by the Prolimatech Megahalems which I have in my system, and my load is around 75C at 4ghz @ 1.28v.

Most air cooled rigs I've seen have similar temps at that clockspeed.......but nothing like what you're saying. Those temps seem more likely on a high end water cooling set up.

Although admittedly it could also be that you keep your computer in a very cold environment..

You have a horrible mount. Back when I was on air my QX9650 @ 4ghz @ 1.36 volts loaded at 60C (quad prime) on my standard TRUE (high teens ambient temp). His temps make sense just fine.

[texasreefer]

I'm at work at the moment, but I have no problems with proving my temps or anything that I post. In fact I have posted some screen shots in the air cooled section, as well as, the water cooled section since my rig is a hybrid setup. BTW the TRUE copper with Noctua fans, and the push/pull mounts, and the LGA 1366 special mounts....set me back almost 250USD. Not exactly pocket change. I can boot up and bench at 4.5 Ghz also, but I don't like having my voltage at 1.44v.

PS - no offense taken...that's what we're all here for, awesome discussions on what nobody at home seems to know anything about.

[Helloworld_98]

You have a horrible mount. Back when I was on air my QX9650 @ 4ghz @ 1.36 volts loaded at 60C (quad prime) on my standard TRUE (high teens ambient temp). His temps make sense just fine.

you can't really compare a Core 2 CPU to a Core i7 Bloomfield when it comes to temps.

[texasreefer]

No offense dude, but I have to see some evidence to believe that. It's not the clockspeed that makes me raise an eye brow, it's the temps.

The True Copper is a great heatsink, but it is equaled and even outperformed in many cases by the Prolimatech Megahalems which I have in my system, and my load is around 75C at 4ghz @ 1.28v.

Most air cooled rigs I've seen have similar temps at that clockspeed.......but nothing like what you're saying. Those temps seem more likely on a high end water cooling set up.

Although admittedly it could also be that you keep your computer in a very cold environment..

Here you go...

http://www.xtremesystems.org/forums/showthread.php?t=231276

[One_Hertz]

you can't really compare a Core 2 CPU to a Core i7 Bloomfield when it comes to temps.

They have identical TDP (both 130) so yes you can.

[dan7777]

vantage no physx...

[Carfax]

You have a horrible mount. Back when I was on air my QX9650 @ 4ghz @ 1.36 volts loaded at 60C (quad prime) on my standard TRUE (high teens ambient temp). His temps make sense just fine.

Nothing is wrong with my mount.

If your ambient temp is low, then thats possible on a Q9650.

But on an i7 with HT enabled, integrated memory controller, integrated PCI-E controller etc, it's most likely not possible.....unless you keep your computer in a very cold environment.

[Carfax]

Here you go...

http://xtremesystems.org/forums/showthread.php?t=231276

Link didn't work......but it doesn't matter.

Sorry to hijack the thread..

[spajdr]

It does, just add www behind http:// .. be a little more creative :)

[Choklate]

my recommendation to u is for now buy the 5870 & wait for nvidia fermi.

[Carfax]

It does, just add www behind http:// .. be a little more creative :)

Hah, I didn't even see that. I just clicked on the link and decided it didn't work.

Anyway, now I know why his temps are so low....or so I think.

He is using PC Probe to measure temps, and not Real Temp or Core temp.

Pc Probe, like other motherboard utilities are notoriously inaccurate when it comes to gauging temperature.

Real temp or Core Temp on the other hand, are very accurate.

I bet if he used Real Temp to measure his loads, his temp would be much higher.

[Blkout]

They have identical TDP (both 130) so yes you can.

No, you can't. Hyperhtreading alone adds 10-12c. It's not myth, it's been proven time and time again by just disabling HT. And as stated above, when you add all the other integrated components in the i7, it's going to run hotter, period. Keep in mind, that while they may have the same TDP, the i7 uses far more of it's TDP threshold due to the nature of it's design.

[wiak]

I would take two 5870.
Reason 1. Fewer watts (one of the 5870 can go into deep sleep mode uses even less when idle)
Reason 2. Fewer GPUs gives better overall performance, 2 (5870) vs 4 (295)
Reason 3. DX11
Reason 4. wiak is awesome!

[texasreefer]

Hah, I didn't even see that. I just clicked on the link and decided it didn't work.

Anyway, now I know why his temps are so low....or so I think.

He is using PC Probe to measure temps, and not Real Temp or Core temp.

Pc Probe, like other motherboard utilities are notoriously inaccurate when it comes to gauging temperature.

Real temp or Core Temp on the other hand, are very accurate.

I bet if he used Real Temp to measure his loads, his temp would be much higher.

Look...here are my temps using almost Every temp app. out there, taken, after a run of 3dMark06. I know that Core temp, Everest, and Real Temp show higher temperatures than PC Probe. That has always been a discussion, and Real Temp has always been the more widely accepted/accurate temp monitor. BUT, PC Probe is directly connected to my Motherboard, and BIOS temps. This is the application that ASUS puts out to measure and ALarm my computer. IF the temps go too high...It shuts off my computer. Just because the Max temp on Real temp shows 74C, and Core Temp shows max of 67C doesn't mean that it spent any significant amount of time at those temperatures. I was running 3dMark06, which included the CPU test, and it may have run up to 75C and then drop back down to 50C. Real temp and Core temp records that temp as the maximum temp, when really the cpu was loading at an average temp that is much lower...say 68C.

Another thing that Real, Core, and Everest will not do for me is monitor/alarm/shutdown if my volatages overload. Also, my GPUs are covered, look at the Optional 1 and Optional 2 temps...these are the temps of one core on each of the 4870X2s. They match up almost to the degree with the Catalyst temps. Those are actual probes that come with the MB and are directly placed on the cores of the videocards. They are accurate enough for me, and my PC is running faster than yours it seems. I use PC Probe because it is the temp monitor that my motherboard uses to alarm and shut down my PC if it gets over heated. If it didn't work Asus would have already put out a different application... I set my shut down temp for my PC at 80C, so if it had hit 80 my computer would have shut down. I'm not testing and runnin trials to make sure my computers safe. I'm just gaming baby! Here scrutinize this pic, so you can tell me what else is wrong with my setup.

http://i52.photobucket.com/albums/g14/texasreefer/TempScreeshot.jpg

[Carfax]

Those are actual probes that come with the MB and are directly placed on the cores of the videocards. They are accurate enough for me, and my PC is running faster than yours it seems. I use PC Probe because it is the temp monitor that my motherboard uses to alarm and shut down my PC if it gets over heated. If it didn't work Asus would have already put out a different application... I set my shut down temp for my PC at 80C, so if it had hit 80 my computer would have shut down.

I just looked at my temps as indicated by the motherboard, and my processor was at 32C..... 12 degrees LESS than what Real Temp and Core Temp are reporting.

Dude, you can do whatever you want, it's your computer. But, if you want ACCURATE temp readings, you can't go off PCprobe.

If you want to see what your true load temps are, use Prime and Real Temp. Prime will fully load the processor which 3DM06 doesn't do, and if I was a betting man, I'd say your temps will exceed 80C easily.

Of course, you could argue that Prime doesn't represent a realistic computing environment, but most people use it as a reference only to find the true limits of their system.

I could hit 4.2ghz on my set up (never tried for 4.4ghz), but the large increase in voltage required for an extra 200mhz just isn't worth it for me......but I don't bench anymore.

[texasreefer]

I just looked at my temps as indicated by the motherboard, and my processor was at 32C..... 12 degrees LESS than what Real Temp and Core Temp are reporting.

Dude, you can do whatever you want, it's your computer. But, if you want ACCURATE temp readings, you can't go off PCprobe.

If you want to see what your true load temps are, use Prime and Real Temp. Prime will fully load the processor which 3DM06 doesn't do, and if I was a betting man, I'd say your temps will exceed 80C easily.

Of course, you could argue that Prime doesn't represent a realistic computing environment, but most people use it as a reference only to find the true limits of their system.

I could hit 4.2ghz on my set up (never tried for 4.4ghz), but the large increase in voltage required for an extra 200mhz just isn't worth it for me......but I don't bench anymore.

ALL of the monitoring systems are different and each has it's own plus' and minus'. I just showed you some real proof of this, the screen shots right there. Core temp (44, 41, 42, 44), Real temp (44, 46, 47, 47), Everest (45, 45, 44, 47), PC Probe (41)....OH yeah....WAY offf The Mark. I told you I ran 3dmark06 not to say I was loading, but rather so that you wouldn't quite predictably say that these weren't true idle temps. So which one is the most accurate? Who knows?

What do you think a representative from Core TEmp would think, or Real TEmp, or even an ASuS representaive would Say......OUr product is better. Just because everyone likes a particular program, doesn't make it more accurate. It's a third party application....like installing a damn APP on an Iphone.

What good are Motherboard readings if you don't use them? Hey broseph, PC Probe is saying my CPU core is at 1.36. My BiOs Setting is at 1.36V....OH, SCHnap maybe my Voltage is really running at 1.5V and I'm just an idiot. I push my system to the point of EXPLODING and any other system i've had....It's called being a PC enthusiast...It's called growing a PaIR. Your not going to be able to see what you PC can do unless you push it.

I don't think I'm overloading my cpu voltage at 1.36v.

You don't bench anymore, wow... :gay: Well i guess coming over to your place is a blast. Running Prime95 and Staring at REal Temp and eating Carmels. Sounds like fun....gtfoh BTW what do you think Prime 95 is.....ITs a Benchmark. Benchmarking isn't only for FRAPS and SCores, it's for gathering information....like temperature and Stability. Other computers run it to show stability, hence, running it is trying to match a BENCHMARK.

Here's what my PC thinks of your PC.... :slapass:

[Chickenfeed]

Sorry to see this thread has been derailed but I just have to add my 10 cents about the initial question once again. I've actually just tested a 295 in my friends 860 system, and I wasn't that impressed. Playing a few games I've been playing at 1920x1200 4xAA MASS on my 5870 feels much nicer for some reason. I've personally used a 295 for a short spurt in my last machine but the Q6600 was stock at the time so I didn't want to make any judgments on it but now after using one in a proper system, I can't say I was that happy. That said I still with my first comment to go with the 5870(s).

As far as this whole temperature thing, can't we just agree and disagree and be gentlemen here guys? No sense fighting over something trivial. If someone does something different than yourself, so be it. I myself personally use Realtemp and load with 8threads of Linpack on max memory load ( one friendly piece of advice, if you are going to post temp info, be so kind as to mention what your load was achieved with as without that tidbit any info is meaningless ) and at my settings get a load of mid to low 70s which is pretty decent if you ask me. I don't use HT as I don't use any apps that show any benefits to having it on, and its only shown to inhibit game performance on the whole so no sense having it on for the temp increase in my eyes. My real world temps are 50-55 playing games and the like and for what Prime is worth, mid 60s small ftt. Just adding some insight to that debate. Anyways please be civil guys!

[Carfax]

OH yeah....WAY offf The Mark. I told you I ran 3dmark06 not to say I was loading, but rather so that you wouldn't quite predictably say that these weren't true idle temps. So which one is the most accurate? Who knows?

Who gives a f*ck about idle temps? We're talking about load temps!

You said your load temps were 58C, which is B.S!

Even after running 3DMark06, your temps climbed to 76C, and 3DMark isn't even CPU intensive.

If you had run Prime for 5 to 10 minutes, it would have been in the 80s for certain.


I push my system to the point of EXPLODING and any other system i've had....It's called being a PC enthusiast...It's called growing a PaIR. Your not going to be able to see what you PC can do unless you push it.

The funny thing is, 3DMark 06 isn't even system intensive. :ROTF:


You don't bench anymore, wow... :gay: Well i guess coming over to your place is a blast.

Yeah, I guess me being a gamer isn't as cool and exciting as you running 3DMark 06 over and over and over again :D


BTW what do you think Prime 95 is.....ITs a Benchmark. Benchmarking isn't only for FRAPS and SCores, it's for gathering information....like temperature and Stability. Other computers run it to show stability, hence, running it is trying to match a BENCHMARK.

You're taking some liberties here.. Anyway, it doesn't matter. The fact is, your claim of a 58C load is a bunch of crap.

You refuted yourself when you posted that screenshot above with a 76C max..


Here's what my PC thinks of your PC.... :slapass:

Well, I may not be an elite OMG bencher like yourself (*snickers*), but I do know that most elite OMG benchers don't use 3D Mark 06 to "load" their system :rofl:

Adios Amigo...

[texasreefer]

nice...... well my vote is 5870

Happy Halloween...