AMD Radeon HD 7970 3 GB

[Mussels]

not to mention with various driver enhancements and bugs, sometimes you do just get weird results.

 

[Kaotik]

Why are you getting so upset?

How the 6870 and 6850 score relative to 7970.
All Res : Equal %
1024x768: 6870 2% lower (the 'wonky' result')
1280x1024 : 6870 2% higher
1680x1050 : 6870 3% higher
1920x1200 : 6870 3% higher
2560x1600 : 6870 2% higher

In one result it is 2% worse than the 6870 when cmpared to the 7970. It could have been a minor glitch and you want him to rebench? It's a review for the highest end single gpu. The low-medium mainstream cards aren't that relevant to the review.

Go here to feel better. http://www.techpowerup.com/reviews/HIS/Radeon_HD_6870/29.html. This is the 6870 original review. It is far more relevant if you want to compare the 6870 and 6850. This review is for the 7970.

Of course newest reviews are always the most relevant, as all the cards are being benched with the said set of drivers & games (unless same set of drivers and games has been used before to bench the said card, which isn't true for this case)

Let's turn those to actual numbers and compare to 1 review ago and the original one too - all have different drivers and games - and of course we won't compare to 7970, but 6870 and 6850 directly

Numbers presented 6870 compared to 6850, 7970 review / 560Ti 448 review / 6990 / 6870 review
All res: 0% faster / 17.9% faster / 14.6% faster / 14.9% faster
1024: 3.1% slower / 14.7% faster / 12.7% faster / 11.1% faster
1280: 1.8% faster / 17.4% faster / 16% faster / 14.9% faster
1680: 5.7% faster / 16.7% faster / 15.6% faster / 14.9% faster
1920: 6% faster / 19.7% faster / 19% faster / 16.3% faster
2560: 4.4% faster / 21.5% faster / 17.9% faster / 16.3% faster

Bonus tip: All reviews between 560Ti 448 and 6870 reviews give similar numbers to those

The relative speed between 2 cards doesn't change notably regardless of which card is being reviewed, only the roundings to nearest full % twist the difference to one way or another a bit, but it doesn't have major impact on it - this is why I've included 6990 in which the 100% mark is even higher than in 7970 review.

Which review doesn't belong, in which review, it's 100% certain that 6870 numbers are plain wrong, which also makes one think if other numbers in the said review are borked, too?

Also, it seems there's some fundamental errors in other reviews too - in 6990 review all 5 resolutions have bigger difference than in 6870 review, but the "all resolutions" difference is smaller

not to mention with various driver enhancements and bugs, sometimes you do just get weird results.

Yes, but when you're talking about 2 cards, which use same chip, of which lower model is using partly disabled chip at lower clocks too, it's impossible, regardless of bugs and whatnot which would be the exact same for both cards, that the lower model of these 2 could outperform the higher model card in tests which are the same for both cards.

edit:
Car comparisons are always fun.
Let's say we have 2 identical Ferraris, which drive straight line - only difference is, the other Ferrari has 1 gear removed and revs limited too. It's impossible that the Ferrari with 1 gear removed and revs limited would outrun the Ferrari with full gears & no rev limits. The full blown Ferrari will keep outrunning the crippled Ferrari no matter if you switch some slowbie driver to fast driver on both cars, and no matter how many faster cars you put to do the same - same applies to 6870 vs 6850.

 

[yogurt_21]

Of course newest reviews are always the most relevant, as all the cards are being benched with the said set of drivers & games (unless same set of drivers and games has been used before to bench the said card, which isn't true for this case)

Let's turn those to actual numbers and compare to 1 review ago and the original one too - all have different drivers and games - and of course we won't compare to 7970, but 6870 and 6850 directly

Numbers presented 6870 compared to 6850, 7970 review / 560Ti 448 review / 6990 / 6870 review
All res: 0% faster / 17.9% faster / 14.6% faster / 14.9% faster
1024: 3.1% slower / 14.7% faster / 12.7% faster / 11.1% faster
1280: 1.8% faster / 17.4% faster / 16% faster / 14.9% faster
1680: 5.7% faster / 16.7% faster / 15.6% faster / 14.9% faster
1920: 6% faster / 19.7% faster / 19% faster / 16.3% faster
2560: 4.4% faster / 21.5% faster / 17.9% faster / 16.3% faster

Bonus tip: All reviews between 560Ti 448 and 6870 reviews give similar numbers to those

The relative speed between 2 cards doesn't change notably regardless of which card is being reviewed, only the roundings to nearest full % twist the difference to one way or another a bit, but it doesn't have major impact on it - this is why I've included 6990 in which the 100% mark is even higher than in 7970 review.

Which review doesn't belong, in which review, it's 100% certain that 6870 numbers are plain wrong, which also makes one think if other numbers in the said review are borked, too?

Also, it seems there's some fundamental errors in other reviews too - in 6990 review all 5 resolutions have bigger difference than in 6870 review, but the "all resolutions" difference is smaller



Yes, but when you're talking about 2 cards, which use same chip, of which lower model is using partly disabled chip at lower clocks too, it's impossible, regardless of bugs and whatnot which would be the exact same for both cards, that the lower model of these 2 could outperform the higher model card in tests which are the same for both cards.

edit:
Car comparisons are always fun.
Let's say we have 2 identical Ferraris, which drive straight line - only difference is, the other Ferrari has 1 gear removed and revs limited too. It's impossible that the Ferrari with 1 gear removed and revs limited would outrun the Ferrari with full gears & no rev limits. The full blown Ferrari will keep outrunning the crippled Ferrari no matter if you switch some slowbie driver to fast driver on both cars, and no matter how many faster cars you put to do the same - same applies to 6870 vs 6850.

math... you needs more classes in it. Each review is re-stated against the current card reviewed. Ie the baseline for comparison changes with each review. This is why percentages are not comperable between reviews and due to changes in drivers and the games benched fps aren't really either.

so sure when the 6870 is the baseline, it shows up as 20% faster on the graph. However should the baseline be say a 5450, it would show as a higher percentage. With a better gpu as the baseline the difference between the cards is going to show as a lower percentage. The difference between other cards on the graphs is only comperable against the baseline. So you must compare each against the baseline to come up with the percentage between the two. Stats doesn't allow you to simply add or subtract percentages of the other cards against each other. That's bad math.

at 1680x1050 the 6870 is actually closer to 10% faster overall compared with the 6850 based on this review. Remember several new bencmarks were introduced in this review and several others were removed. You simply cannot use the other numbers to justify why you think this review is off.

if you're really interested in the how and why of all this I'm sure there's plenty of course offerings in your area on stats and percentages.

 

[Kaotik]

math... you needs more classes in it. Each review is re-stated against the current card reviewed. Ie the baseline for comparison changes with each review. This is why percentages are not comperable between reviews and due to changes in drivers and the games benched fps aren't really either.

so sure when the 6870 is the baseline, it shows up as 20% faster on the graph. However should the baseline be say a 5450, it would show as a higher percentage. With a better gpu as the baseline the difference between the cards is going to show as a lower percentage. The difference between other cards on the graphs is only comperable against the baseline. So you must compare each against the baseline to come up with the percentage between the two. Stats doesn't allow you to simply add or subtract percentages of the other cards against each other. That's bad math.

at 1680x1050 the 6870 is actually closer to 10% faster overall compared with the 6850 based on this review. Remember several new bencmarks were introduced in this review and several others were removed. You simply cannot use the other numbers to justify why you think this review is off.

if you're really interested in the how and why of all this I'm sure there's plenty of course offerings in your area on stats and percentages.

Wait what, I need math classes? Did you even read what I wrote?

If you bothered checking my numbers, you'd see the numbers are correct - they're all counted as how many %'s is the 6870's number from 6850's - the baseline is irrelevant for that, moving the 100% mark only causes some rounding errors. I used 6990 review too just to illustrate this, and could use 5450 review too.

They're all counted as how many percents faster (or slower in 1 case) 6870 is compared to 6850 - 11.1% faster is the same regardless if of what you have as baseline, 6850 27 vs 6870 30, 6850 90 6870 100, 6850 216 6870 240 - they're all the same.

I know several games were changed for this review, and that can of course affect the difference between 2 cards a bit, but it doesn't do what happened in this review, it doesn't put a card ~15-20% faster on every occasion before to be around 5% faster to actually being slower just because you switch games

 

[nt300]

I wonder how the HD 7850 and 7870 will perform and how much they will cost?

 

[Mussels]

I wonder how the HD 7850 and 7870 will perform and how much they will cost?

i'm more interested in their power to performance ratio, they could turn out to be extremely fast for their power consumption, which is a massive win for ITX/mATX system builders.

 

[W1zzard]

problem found why hd 6870 sucks so much since the rebench:

result of the damage -> card running at pcie x2 (that's 2 lanes vs. 16 it should run at)

off to find another hd 6870

 

[eidairaman1]

hmmm I must admit I was expecting a bit more from this chip, plus for the price they are asking there is no way I am trading in my 6950. If they priced it at £300 I would be sorely tempted but £450 is ridiculus. Once Nvidia release their 28nm cards im sure we will see a huge price drop. It must be AMD trying to make up for the Bulldozer fail and claw back more profits

Incorrect. The 7000 Series had a precise launch and met or exceeded performance expectations, Marketing of this model series hasnt been exaggerated by the marketing dept. FYI Bulldozer is far from a failure because people and businesses are buying them.

 

[cadaveca]

Ouch. Good catch though, W1zz!

 

[crazyeyesreaper]

yup that definitely a bummer and explains the discrepancies in performance.

 

[pantherx12]

problem found why hd 6870 sucks so much since the rebench:

http://img.techpowerup.org/111224/Capture244.jpg

result of the damage -> card running at pcie x2 (that's 2 lanes vs. 16 it should run at)

off to find another hd 6870

One day I hope those get embedded in the PCB it's self rather than surface mounted.

Knocked so many off them off over the years.

 

[eidairaman1]

Never knocked smds off. Ive shorted ram out n bent a cpu pin but thats pretty much it

 

[twicksisted]

wow 3gb of gfx ram, wonder how crap it performs on a 32bit OS!

 

[W1zzard]

Never knocked smds off. Ive shorted ram out n bent a cpu pin but thats pretty much it

don't store your electronics like this and you will be fine

 

[twicksisted]

don't store your electronics like this and you will be fine

http://img.techpowerup.org/111224/Capture245.jpg

"Boxing" day at Wizz's house

 

[cadaveca]

don't store your electronics like this and you will be fine

http://img.techpowerup.org/111224/Capture245.jpg

You're doing better than I am At least some of yours are seperated by carboard.:

 

[acperience7]

I was really looking forward to seeing their liquid chamber tech on these cards. Other than that I'm really liking the performance numbers, but I think I'll be able to get by on my 5870's for a while.

 

[twicksisted]

but on a serious note, what happens if you try and use one of these on a 32bit OS?
If a 32bit OS can only address 4gb of ram in total that doesent leave much left for the gfx memory to be duplicated in the system memory or has this since changed in Win7?

 

[arnoo1]

don't store your electronics like this and you will be fine

http://img.techpowerup.org/111224/Capture245.jpg

please donate me one

 

[W1zzard]

At least some of yours are seperated by carboard

i added the cardboard today after discovering the damaged hd 6870. but then was too lazy to do it for the other boxes

but on a serious note, what happens if you try and use one of these on a 32bit OS?

the card will only allocate a 256 or 512 mb window in the cpu address space (i actually checked this with a hd 7970, but forgot which of the two it was)

 

[MxPhenom 216]

If the 7870 has GTX570 performance but low power consumption. it might be my card! maybe 2 of them.

okay yep. Might get a 7870

http://wccftech.com/amd-launching-r...n-xtpro-february-2012-pricing-specs-detailed/

 

[Kaotik]

wow 3gb of gfx ram, wonder how crap it performs on a 32bit OS!

I wouldn't worry about the cards performance so much - i'd be more worried how on earth the OS itself would perform with the leftover address space

 

[Kaotik]

the card will only allocate a 256 or 512 mb window in the cpu address space (i actually checked this with a hd 7970, but forgot which of the two it was)

This can't be right
It has to map everything it's going to ever use to the address space there is, which is exactly 4GB in 32bit system, since PAE hasn't been supported since.. was it XP SP2?

Pushing the Limits of Windows: Physical Memory - Mark's Blog - Site Home - TechNet Blogs

The consumption of memory addresses below 4GB can be drastic on high-end gaming systems with large video cards. For example, I purchased one from a boutique gaming rig company that came with 4GB of RAM and two 1GB video cards. I hadn't specified the OS version and assumed that they'd put 64-bit Vista on it, but it came with the 32-bit version and as a result only 2.2GB of the memory was accessible by Windows. You can see a giant memory hole from 8FEF0000 to FFFFFFFF in this Meminfo output from the system after I installed 64-bit Windows:

The system memory that is reported in the System Information dialog box in Windows Vista is less than you expect if 4 GB of RAM is installed

Various devices in a typical computer require memory-mapped access. This is known as memory-mapped I/O (MMIO). For the MMIO space to be available to 32-bit operating systems, the MMIO space must reside within the first 4 GB of address space.

For example, if you have a video card that has 256 MB of onboard memory, that memory must be mapped within the first 4 GB of address space. If 4 GB of system memory is already installed, part of that address space must be reserved by the graphics memory mapping. Graphics memory mapping overwrites a part of the system memory. These conditions reduce the total amount of system memory that is available to the operating system.

The reduction in available system memory depends on the devices that are installed in the computer. However, to avoid potential driver compatibility issues, the 32-bit versions of Windows Vista limit the total available memory to 3.12 GB. See the "More information" section for information about potential driver compatibility issues.

If a computer has many installed devices, the available memory may be reduced to 3 GB or less. However, the maximum memory available in 32-bit versions of Windows Vista is typically 3.12 GB.

Memory Limits for Windows Releases

How graphics cards and other devices affect memory limits

Devices have to map their memory below 4 GB for compatibility with non-PAE-aware Windows releases. Therefore, if the system has 4GB of RAM, some of it is either disabled or is remapped above 4GB by the BIOS. If the memory is remapped, X64 Windows can use this memory. X86 client versions of Windows don’t support physical memory above the 4GB mark, so they can’t access these remapped regions. Any X64 Windows or X86 Server release can.
X86 client versions with PAE enabled do have a usable 37-bit (128 GB) physical address space. The limit that these versions impose is the highest permitted physical RAM address, not the size of the IO space. That means PAE-aware drivers can actually use physical space above 4 GB if they want. For example, drivers could map the “lost” memory regions located above 4 GB and expose this memory as a RAM disk.

 

[Mussels]

Kaotik: thats not how much is usable max, its how much is reserved, (minimum) per card.

 

[W1zzard]

finished rebenching the hd 6870 and updated the summary graphs