• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

AMD Richland A10-6800K APU reviews.

humbug said:
Its more than 5% ^^^



yup.

http://www.hardwarecanucks.com/foru...-review-a10-6800k-a10-6700-benchmarked-4.html

3DMark11 +8%

http://www.hardwarecanucks.com/foru...review-a10-6800k-a10-6700-benchmarked-19.html

Human revolution +16%

Dirt3 +7%

Skyrim +9%

http://www.hardwarecanucks.com/foru...review-a10-6800k-a10-6700-benchmarked-20.html

I'm sure running higher Memory speed vs Trinity would help a bit, but clearly that's far from all that's going on here, HC are running 1600Mhz on both Richland and Trinity (its apples for apples) and yet even in the CPU dependant Dirt 3 and Skyrim gets 7 to 10%, Human revolution is 16%

The iGPU its self has got a boost. quite a significant one at that.
Click to expand...

It really depends on what reviews you read. Xbitlabs has quite underwhelming gaming improvement from Trinity to Richland;
http://www.xbitlabs.com/articles/cpu/display/amd-a10-6800k_6.html#sect0

In that ^ generally the gains are a little less than the 5.5% clock speed increase, which is what you would expect if there had only been a clock boost.

EDIT:

In fact, HC seems to be about the only set of reviews which shows such gains. Even Overclockers.com, figures (where they used faster ram for Richland) didn't show such a big gain; +5.5% for Dirt, which is less than the HC one which used, apparently, the same ram speed.
http://www.overclockers.com/amd-richland-a106800k-apu-review/
 
Last edited:
cheesyboy said:
It really depends on what reviews you read. Xbitlabs has quite underwhelming gaming improvement from Trinity to Richland;
http://www.xbitlabs.com/articles/cpu/display/amd-a10-6800k_6.html#sect0

In that ^ generally the gains are a little less than the 5.5% clock speed increase, which is what you would expect if there had only been a clock boost.

EDIT:

In fact, HC seems to be about the only set of reviews which shows such gains. Even Overclockers.com, figures (where they used faster ram for Richland) didn't show such a big gain; +5.5% for Dirt, which is less than the HC one which used, apparently, the same ram speed.
http://www.overclockers.com/amd-richland-a106800k-apu-review/
Click to expand...

You determined, i'll give you that...

I however am not, its not that important to me, moving on....
 
FWIW the Trinty A10 4600m in my laptop mysteriously outpaces every review I've seen of them, using generic 1600mhz RAM. Looked similar for a 5800k build I did for someone but didn't have enough time to play about with it
 
d_brennen said:
FWIW the Trinty A10 4600m in my laptop mysteriously outpaces every review I've seen of them, using generic 1600mhz RAM. Looked similar for a 5800k build I did for someone but didn't have enough time to play about with it
Click to expand...


The same is not unusual for Discrete GPU's, when I look around a lot of them don't match up what I get when I replicate with they did, especially TPU and Anand its as if they take the results they got, and then subtract 5 or 10%.

But its also becoming increasingly difficult to compare as those reviewers no longer say what maps they are benching, so you can't replicate it.

For me if they don't say how they bench their result is void.

A proper 'independent' reviewer has no problem detailing how they got to their results so it can be checked, and they used to, now its all commercialised. big reviewers are run in just the same way big business is run, taking money from interested parties in return for certain requests.

I trust much smaller far less known reviewers far more than I do the main ones.
 
humbug said:
The same is not unusual for Discrete GPU's, when I look around a lot of them don't match up what I get when I replicate with they did, especially TPU and Anand its as if they take the results they got, and then subtract 5 or 10%.
Click to expand...
Driver improvements;
http://www.hardocp.com/article/2013...card_driver_performance_review/6#.UbBDnECsiSo

Reviews are taken when the product is new, and the drivers not yet matured.

Tom's Hardware's Richland review is up;
http://www.tomshardware.com/reviews/a10-6700-a10-6800k-richland-review,3528.html
 
cheesyboy said:
Driver improvements;
http://www.hardocp.com/article/2013...card_driver_performance_review/6#.UbBDnECsiSo

Reviews are taken when the product is new, and the drivers not yet matured.

Tom's Hardware's Richland review is up;
http://www.tomshardware.com/reviews/a10-6700-a10-6800k-richland-review,3528.html
Click to expand...


I always use the same driver as the review.

Does CF not work on Richland? every game they tested with the APU and a 6670 gave no performance improvement, its as if its not even there.
 
VLIW 5 the 5XXX is, not 4.

I'm pretty confident the 6670 doesn't run the improved shader cores seen in the 69XX, and instead runs the same as the 5670.
It explains how the Richland APU is performing over Trinity with the same shader count.
 
Last edited:
Martini1991 said:
VLIW 5 the 5XXX is, not 4.

I'm pretty confident the 6670 doesn't run the improved shader cores seen in the 69XX, and instead runs the same as the 5670.
It explains how the Richland APU is performing over Trinity with the same shader count.
Click to expand...



So what your saying is the 6900 runs better cores than the 66/700 series which is the same as the 56/700 series, and Trinity is running the 56/700 series while Richland runs the better 6900 series.
 
humbug said:
Yes :)

The 6670 has 100 more SP's and GDDR5, considering they are both based on the same VLIW4 architecture its does not compute does it?
Click to expand...

I think it must be down to memory bandwidth. They're using a DDR3 6670 (rather than a GDDR5) which has bandwidth of around 29GB/s; 2133 mhz ram with the IGP will give around 34GB/s (30GB/s for 1866mhz)
 
Last edited:
Pretty much.
Although I'm pretty sure that Wiki is wrong, the 68XX were running the improved shaders (Or at least something different to that of the 5XXX)
The 69XX shaders were kind of a half way thing to GCN, they were improved over the 58XX shaders, hence why a 1,600 shader 5870 gets beaten by the 1536 shader 6970.

cheesyboy said:
I think it must be down to memory bandwidth. They're using a DDR3 6670 (rather than a GDDR5) which has bandwidth of around 29GB/s; 2133 mhz ram with the IGP will give around 34GB/s (30GB/s for 1866mhz)
Click to expand...


That could also be the reason, assuming it is the GDDR3 6670, not the GDDR5 as mentioned (Hence my explanation for why it could be better than that part)
 
Last edited:
Both Trinity and Richland use VLIW4 shaders.

Edit!!

BTW,the bandwidth for the A10 is shared by BOTH the CPU and IGP,so the IGP will probably still be receiving less bandwidth than an HD6670 GDDR3 I suspect.
 
Last edited:
CAT-THE-FIFTH said:
Both Trinity and Richland use VLIW4 shaders.

Edit!!

BTW,the bandwidth for the A10 is shared by BOTH the CPU and IGP,so the IGP will probably still be receiving less bandwidth than an HD6670 GDDR3 I suspect.
Click to expand...

I've not tried the calculations until now, but wouldn't total memory bandwidth of two 2133mhz modules be 68GB/s? suggesting the IGP only soaks up up to half of what's available?

bandwidth calculation;
2133 x 2 x 64 x 2 = 68,256 MB/s = 68 GB/s
mhz x DDR x bits x dual modules = bandwidth

(my 34GB/s was based on quoted bandwidth rates for IGP)

Or have I got something wrong?
 
cheesyboy said:
I've not tried the calculations until now, but wouldn't total memory bandwidth of two 2133mhz modules be 68GB/s? suggesting the IGP only soaks up up to half of what's available?

bandwidth calculation;
2133 x 2 x 64 x 2 = 68,256 MB/s = 68 GB/s
mhz x DDR x bits x dual modules = bandwidth

(my 34GB/s was based on quoted bandwidth rates for IGP)

Or have I got something wrong?
Click to expand...

I don't think the graphics cards have dual channel memory controllers,so you might have a point there.
 
Martini1991 said:
http://forums.overclockers.co.uk/showthread.php?t=18447230&highlight=Trinity

That's launch at circa 100.
So it's over 25% premium in price. At least from Overclockers.

That said, even at a place where the A10 68 is cheaper, the A10 58 was also cheaper, so it's still a 25% premium which ever way it's sliced :p
Click to expand...
So is it going to be how it is then? First looking at the GTX780 and now the A10-6800K...is performance improvement with moving gen (without price increase) a thing of the pass? We are expected to pay a price premium if a next gen thing is faster than the last gen thing despite they are meant to be faster in the first place? :rolleyes:

I mean with the price of the GTX780 (£550) and the A10-6800K (£125) at, had they been launched alongside with the GTX680 (£400-£420) and the A10-5800K (sub £100), their prices would still had not been out of place.
 
Last edited:
You must log in or register to reply here.
Back
Top Bottom