vivster said:
If they decided against the ESRAM and put in the same number of CUs the difference would've only been in the single digits due to the slower RAM. The ESRAM isn't even close in making up the power difference lost due to the missing CUs. |
Ofcourse ESRAM can't make up for performance of the lost CUs, but if they had 18 CUs with no ESRAM & DDR3 you would still expect much lower performance than the PS4.
If DDR3 only made a difference of single digits then both companies would just use that instead of using the GDDR5/ESRAM options.
I was trying to find some like-for-like comparisons of DDR3 vs GDDR5. Here's some.
http://www.redgamingtech.com/ps4-vs-xbox-one-gddr5-vs-ddr3-latency/

http://tech.firstpost.com/reviews/his-hd-6670-1gb-ddr3-graphics-card-20680.html#show

Here's another on more similar hardware to PS4/XB1 (HD 7750).
| Scene | Card | 720p (avg) |
| Dark Tower | AMD HD 7750 1GB GDDR5 | 53.12 |
| AMD HD 7750 2GB GDDR3 | 34.15 | |
| Desert Sunrise | AMD HD 7750 1GB GDDR5 | 72.03 |
| AMD HD 7750 2GB GDDR3 | 50.41 | |
| Concrete Jungle | AMD HD 7750 1GB GDDR5 | 41.64 |
| AMD HD 7750 2GB GDDR3 | 27.04 |
And this:
Basically as soon as they picked DDR3 with any combo is worse than GDDR5. Lower # of CUs with ESRAM is slower. Same number of CUs without ESRAM is slower.








