Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Thief^]

I have repeatedly said to look at the (few, admittedly) stock models in the roundups.  I'm done

There aren't any stock models at all in your latest link...

[gimlet]

Sapphire Dual-X 100365L R9 270 2 GB
HIS iPower IceQ X² Turbo H270QMT2G2M R9 270 2 GB
Sure look like "stock models" to me - why do you think they're not?  AFAIK they were 2 of the first models of the R9-270 released.

[Evaris]

Sapphire Dual-X 100365L R9 270 2 GB
HIS iPower IceQ X² Turbo H270QMT2G2M R9 270 2 GB
Sure look like "stock models" to me - why do you think they're not?  AFAIK they were 2 of the first models of the R9-270 released.

Custom coolers, different RAM chips, the HIS comes with a factory overclock, should I go on?

[gimlet]

"Custom coolers and different RAM chips" - I thought the claim was that those made no difference at all.

Geez, so only if it's completely identical to the reference card do you consider it a "stock model"?   Well, yeah, then all those are gonna perform pretty similar...

[Evaris]

"Custom coolers and different RAM chips" - I thought the claim was that those made no difference at all.

Geez, so only if it's completely identical to the reference card do you consider it a "stock model"?   Well, yeah, then all those are gonna perform pretty similar...

They make a difference, regardless of what others might tell you.  The R9 series utilize a boost function - therefore a better cooler allows it to stay at boost frequencies longer, regardless of any other overclocking. 

Different RAM may improve overclocking, should you undertake it. 

So yes, there is a performance difference, likely between 5-15% in comparison to a stock card.  (which is sizable and not what could be considered "similar" to stock.)

Also the definition of a "stock model" of a graphics card is one which follows the reference card... so yeah.

[Thief^]

Sapphire Dual-X 100365L R9 270 2 GB
HIS iPower IceQ X² Turbo H270QMT2G2M R9 270 2 GB
Sure look like "stock models" to me - why do you think they're not?  AFAIK they were 2 of the first models of the R9-270 released.

The reference clock of a 270 is 925 MHz, the sapphire is 945 and the HIS "Turbo" is 975. Neither is stock clocked. (Also, a relatively pointless 2% and 5%, good luck noticing that difference in games, it equates to approx 1 and 3 extra fps respectively at ~60 fps)
On the their website the sapphire is listed as "SAPPHIRE DUAL-X R9 270 2GB GDDR5 WITH BOOST & OC" - http://www.sapphiretech.com/presentation/product/?cid=1&gid=3&sgid=1227&pid=2039&psn=&lid=1&leg=0


I've taken the time to look at that last link gimlet posted in detail, and although it's curiously missing any benchmarks or conclusions whatsoever, it does have a nice test of the cards' ability to hold their clock under load: http://www.tomshardware.com/reviews/radeon-r9-270-benchmark-review,3769-2.html
Specifically the "Sapphire Vapor-X 100364VXL Radeon R9 270X 2 GB" can't maintain 1100, dropping to 1050 repeatedly, and the "Asus DirectCU II R9270X-DC2T-4GD5 R9 270X 4 GB" can't maintain 1120, also dropping to 1050. The "HIS IceQ X² Turbo Boost Clock H270XQMT2G2 R9 270X 2 GB" on the other hand can maintain 1140 perfectly happily.
Interestingly this doesn't seem to be caused by the cooler, officially the R9 series downclocks at 95C, none of the cards on test run at over 80C, and neither of the cards that can't hold their clock speed is the hottest card! The "MSI Radeon R9 270X Gaming 4 GB" runs at 77C (vs the 73C and 71C of the Sapphire and Asus cards), but still maintains its 1120 MHz clock rate happily!

If anything, this is evidence that the different coolers make no difference what-so-ever to the performance of the card.

[Evaris]

Sapphire Dual-X 100365L R9 270 2 GB
HIS iPower IceQ X² Turbo H270QMT2G2M R9 270 2 GB
Sure look like "stock models" to me - why do you think they're not?  AFAIK they were 2 of the first models of the R9-270 released.

The reference clock of a 270 is 925 MHz, the sapphire is 945 and the HIS "Turbo" is 975. Neither is stock clocked. (Also, a relatively pointless 2% and 5%, good luck noticing that difference in games, it equates to approx 1 and 3 extra fps respectively at ~60 fps)
On the their website the sapphire is listed as "SAPPHIRE DUAL-X R9 270 2GB GDDR5 WITH BOOST & OC" - http://www.sapphiretech.com/presentation/product/?cid=1&gid=3&sgid=1227&pid=2039&psn=&lid=1&leg=0


I've taken the time to look at that last link gimlet posted in detail, and although it's curiously missing any benchmarks or conclusions whatsoever, it does have a nice test of the cards' ability to hold their clock under load: http://www.tomshardware.com/reviews/radeon-r9-270-benchmark-review,3769-2.html
Specifically the "Sapphire Vapor-X 100364VXL Radeon R9 270X 2 GB" can't maintain 1100, dropping to 1050 repeatedly, and the "Asus DirectCU II R9270X-DC2T-4GD5 R9 270X 4 GB" can't maintain 1120, also dropping to 1050. The "HIS IceQ X² Turbo Boost Clock H270XQMT2G2 R9 270X 2 GB" on the other hand can maintain 1140 perfectly happily.
Interestingly this doesn't seem to be caused by the cooler, officially the R9 series downclocks at 95C, none of the cards on test run at over 80C, and neither of the cards that can't hold their clock speed is the hottest card! The "MSI Radeon R9 270X Gaming 4 GB" runs at 77C (vs the 73C and 71C of the Sapphire and Asus cards), but still maintains its 1120 MHz clock rate happily!

If anything, this is evidence that the different coolers make no difference what-so-ever to the performance of the card.

Temperatures obviously do matter for the OC of the cards.  Furthermore, take into account the decibel levels of the different coolers, as well as the fact that a cooler running card will last longer. 

But sure, feel free to make an assertion contrary to evidence and the opinion of the hardware enthusiast community as a whole.  (I am a member of Techpowerup!'s forums as well as HardOCP... I know what the general opinion of aftermarket and different OEM coolers are.) 

[Thief^]

If temperature matters so much, explain how with the same gpu the Asus card can't maintain its 1120 MHz boost at 71C but the MSI maintains 1120 MHz even at 77C? How come the lower-temperature one (theoretically meaning a better cooler) is running worse? Will the card throttle itself back for no apparent reason even if you manually overclock it? The answer would seem to be that they have different thresholds for their "boost", but try to find that documented anywhere, even though it would seem to be rather important to its performance (if you care about a few percent).

As for a cooler card lasting longer... It's completely unmeasurable and incalculable, and therefore meaningless! Answer me this: If I take a stock 270X (which I found a review saying it runs at 76C under load) and stick it in a shitty case with terrible airflow to the point it will run at 95C under load (the throttle-back threshold for the R9) how long will it last? 1 day less? Half as long? Only 1 day?
Choose the card with the longest guarantee if you're after long life. If that card happens to have a better cooler (it may well) win-win!

I used to overclock, back when you could do things like alter the FSB on an 1866MHz AMD 2500+ cpu from 166MHz to 200MHz to transform it into a top-of-the-range 2200MHz 3200+ cpu (complete with name, weirdly) for a 20% boost, or get an XP-m chip and push it even higher (e.g. I have records of pushing a 2500-m to 2400MHz or a 2400-m to 2580 MHz, which was faster than any cpu that could be bought at the time, 17% faster than even the top-of-the range 3200+). I have benchmark records of running an nVidia 8800 GTS at 650MHz core instead of 500, a 30% overclock. That kind of stuff mattered.
These days, thanks to the extended last console generation and everything being optimised for those... even my aging GTX 285 runs everything, so why bother? I'd bet even the R9 270 at stock outperforms the xbox one.

[Evaris]

If temperature matters so much, explain how with the same gpu the Asus card can't maintain its 1120 MHz boost at 71C but the MSI maintains 1120 MHz even at 77C? How come the lower-temperature one (theoretically meaning a better cooler) is running worse? Will the card throttle itself back for no apparent reason even if you manually overclock it? The answer would seem to be that they have different thresholds for their "boost", but try to find that documented anywhere, even though it would seem to be rather important to its performance (if you care about a few percent).

As for a cooler card lasting longer... It's completely unmeasurable and incalculable, and therefore meaningless! Answer me this: If I take a stock 270X (which I found a review saying it runs at 76C under load) and stick it in a shitty case with terrible airflow to the point it will run at 95C under load (the throttle-back threshold for the R9) how long will it last? 1 day less? Half as long? Only 1 day?
Choose the card with the longest guarantee if you're after long life. If that card happens to have a better cooler (it may well) win-win!

I used to overclock, back when you could do things like alter the FSB on an 1866MHz AMD 2500+ cpu from 166MHz to 200MHz to transform it into a top-of-the-range 2200MHz 3200+ cpu (complete with name, weirdly) for a 20% boost, or get an XP-m chip and push it even higher (e.g. I have records of pushing a 2500-m to 2400MHz or a 2400-m to 2580 MHz, which was faster than any cpu that could be bought at the time, 17% faster than even the top-of-the range 3200+). I have benchmark records of running an nVidia 8800 GTS at 650MHz core instead of 500, a 30% overclock. That kind of stuff mattered.
These days, thanks to the extended last console generation and everything being optimised for those... even my aging GTX 285 runs everything, so why bother? I'd bet even the R9 270 at stock outperforms the xbox one.

1.  You also need to consider the circuit layout and power phases on aftermarket cards - temperature is not the only thing which matters, although it does.

2.  While I'm not certain, I believe the estimate for silicon operating above temperatures between 80-100*C is 1*C = about 2500 hours of operational time, if I remember correctly, granting that most silicon chips have an estimated lifespan of 100,000 hours or so of continued use. 

3.  Guess what, power still matters.  Sure, you can get console-quality graphics off a stock 270x, given the PS4 has a 260x equivalent on-board and consoles are better optimized.  However, I like to keep in mind you can get better graphics still if you want them, and PC offers higher graphics to those who have a PC capable of them.  If you want better graphics, or a smoother framerate - more power gives you those options.  And new games are coming out all the time - just because things are okay for current generation games doesn't mean it will be suitable for games later this year or next year.  New generation games are starting to permeate as it is, even in MMOs - take a look at Black Desert, for example.  Sure, your 285 might meet the minimum requirements.  But are you sure the OP is looking for something which only meets the minimum, or might the OP try to get a bit more performance for the dollar? 

[Thief^]

I didn't even mean minimum
I run most things on high/max. It's supremely disappointing that that's even possible. It's roughly between the R7 250 / 250X graphics cards in performance, so you can still buy significantly less powerful cards. GPU performance seems to not be increasing as fast as it once was.

It's theoretically 75% the power of the xbox one's gpu, so a large majority of games will run acceptably on it for the forseeable future... and it's 5 years old

[Evaris]

I didn't even mean minimum
I run most things on high/max. It's supremely disappointing that that's even possible. It's roughly between the R7 250 / 250X graphics cards in performance. GPU performance seems to not be increasing as fast as it once was.

It's theoretically 75% the power of the xbox one's gpu, so a large majority of games will run acceptably on it for the forseeable future... and it's 5 years old

Well it's what you get for buying high-end graphics cards. 

Also I highly doubt you run games like Crysis 3 on high/max, but that is besides the point.

Graphics cards, and silicon processors in general, are improving at a slower rate than they once did, this is true.  For the last five years or so, top-end performance gains have only been 15-30% / year with each generation, and graphics card generations have been moving slower and slower, alongside delays in node shrinks. 

[Thief^]

True, but crysis is a glorious exception to the general mediocrity. This benchmark of the (actually less powerful) R7 250 suggests I might have to turn it all the way down to medium: http://www.expertreviews.co.uk/graphics-cards/1305007/amd-radeon-r7-250.

EDIT: Oh hey, here's an interesting point: http://hexus.net/tech/reviews/graphics/65977-sapphire-radeon-r7-250-dual-graphics-mantle/?page=3
"the A10-7850K's onboard R7 graphics are actually a tiny bit faster than the R7 250 card."

I wonder how long it will be before discrete GPUs are obsoleted?

[Evaris]

I wonder how long it will be before discrete GPUs are obsoleted?

Likely not until we are utilizing photon or quantum processors, where latency becomes a greater bottleneck than the thermal bottleneck.  That said, for anything less than enthusiast level cards, we're really already at it, with the R9 270's required to see any large advantage over the top end integrated cards presently.   

Though as another point, 4k monitors are coming down in price, and are in the $350 range for a low-end 4k TV/monitor, and in the $550 range for a good one.  Given a year or two and 4k will be in the $200 range of mass adaptation - and we see where graphics power need will see the largest increase for the most part, given the 4 times greater pixel count.

[Thief^]

I must admit, I'm looking forward to the day I can get a 4k monitor, it's been too long since we've had worthwhile DPI.

I used to use a 1600x1200 resolution 19" CRT (most likely 18" viewable), which is ~111 dpi. My current work monitor is a 27" 1080p LCD, which is only ~81 dpi (and actually fewer vertical pixels, and only 8% more pixels in total).

Free/G -sync is another beautiful thing, but why no-one thought of it during the CRT era (when the graphics card *did* control the monitor refresh rate) I have no idea.

[Tellemurius]

I must admit, I'm looking forward to the day I can get a 4k monitor, it's been too long since we've had worthwhile DPI.

I used to use a 1600x1200 resolution 19" CRT (most likely 18" viewable), which is ~111 dpi. My current work monitor is a 27" 1080p LCD, which is only ~81 dpi (and actually fewer vertical pixels, and only 8% more pixels in total).

Free/G -sync is another beautiful thing, but why no-one thought of it during the CRT era (when the graphics card *did* control the monitor refresh rate) I have no idea.

They didn't worry about it much back then cause CRT always had manual control over the Refresh rate plus i doubt people worry much about Doom 2 running 60fps