• Welcome to TechPowerUp Forums, Guest! Please check out our forum guidelines for info related to our community.
  • The forums have been upgraded with support for dark mode. By default it will follow the setting on your system/browser. You may override it by scrolling to the end of the page and clicking the gears icon.

Sapphire HD 4890 1 GB GDDR5

W1zzard

Administrator
Staff member
Joined
May 14, 2004
Messages
28,899 (3.74/day)
Processor Ryzen 7 5700X
Memory 48 GB
Video Card(s) RTX 4080
Storage 2x HDD RAID 1, 3x M.2 NVMe
Display(s) 30" 2560x1600 + 19" 1280x1024
Software Windows 10 64-bit
Sapphire's new HD 4890 comes with 1 GB of GDDR5 memory and yet unseen clock speeds offering much better performance at a competitive price. Will that be enough to beat NVIDIA's new GeForce GTX 275 ?

Show full review
 
Last edited:
Any possible way we could see GTA4 .....please.....please....please
 
good review as always, love your board components analysis and OC and power consumption section.

could you please add info about how you bench these games , like Clear sky has a benchmark , do you use that or how else do you bench ? and other games too. thank you
 
wow what a loud card.
 
The idle power consumption is alarming.... Not sure why it is so high, especially when the GTX260 is well above it under load.


Perhaps the high idle power consumption is linked to the idle fan speed, which may be calibrated for dust-ridden poorly ventilated cases, to avoid problems that some have with overheating.
 
So it's about 10 frames a sec faster on most games avg.

If I took that review right, and really doesn't eat that much more voltage over the 70.

A few of the games the difference was way off, and that to me looked like driver bugs more than it being that much more powerful of a card. IDK if I am wrong about that or not.
 
Curiously, TPU's sample is "only" 850MHz and in all scenarios it - or rather, the whole rig - consumed more or around equal amount of power than a HD4870.
Now, xbitlabs has a 900MHz sample and here's what they found when they measured the card's consumption (card only):
xbit_chart_4890_power.png

HD4890 @ 900MHz consumed significantly less power than HD4870, in all scenarios.

The GDDR5 memory chips are made by Qimonda and carry the model number IDGV1G-05A1F1C-40X. With a latency of 1.0 ns, they are specified to run at 1000 MHz.
1.0ns is the cycle time of the chip, not a latency.
This is AMD's new RV790 GPU. It is made in a 55 nm process at TSMC Taiwan with 956 million transistors. Please note that the die size is slightly increased from the RV770, the exact reason for that is unknown.
RV790 has 959 million transistors.
AMD has put two VT1165 voltage controllers on their cards. One is responsible for the VGPU core voltage and the second one controls the memory voltage.
Memory vDD is 1 phase, vDDQ is also 1 phase. vGPU is 3 phases (CPL).
What are those 2 phases for (that with the Pulse inductor pack)? And what controls them?
 
Last edited:
Please add Assasins Creed and GTA 4 to the benchmarks.

(and maybe Take out Prey)
 
at the end...


like a PLUS you can add the "possible" support for HAVOC
 
Awesome review, but you did get lucky with this sample, just not the power color one.
 
In another review of an asus card they hit 990mhz core so maybe the asus cards are the ones to get.
 
Any chance of some GPU-Z pics (Sensor Tab) which show the VDDC during 2D and 3D states? I'm interested in how aggressive the BIOSes on these cards are set, with regards to voltage. That's the only way I can figure these HD4890s are using less power in 2D IDLE than comparable HD4870 512MB/1GB cards...
 
Great review as always (Wizz's ones are a long time favorite of mine for readability and comprehensiveness).

What I am missing here, however, are the results of the HD4870 1Gb version.
 
another fine review by the wizz
 
Interesting. I would have liked to see a preview of the volt-mod software, but thorough as always. I want to know how W1z gets those scores in 06 with a 295 lol
 
excellent review W1z, as always. pity it doesn't bring more to the table, and a pity your samples didnt overclock so well.

A lesson to be learned, don't buy a 4890 EXPECTING 1ghz core, you may be disappointed, maybe not but its a gamble.
 
I'm just shocked that 4890's now run Crysis better than GTX 285's. The world has been stood on its head.
 
Didn't AMD directly state that the shaders have been optimized for increased efficiency? As such, I'm not sure why the review claims nothing other than clock speed has changed and completely dismisses the increased die size. This is clearly not the same chip.
 
Didn't AMD directly state that the shaders have been optimized for increased efficiency? As such, I'm not sure why the review claims nothing other than clock speed has changed and completely dismisses the increased die size. This is clearly not the same chip.

source?
 
I saw it on XS before it went down. When it comes back online it should still be there. Proof aside (for the moment), we do know the pinout changed and the die size increased, so the shaders being changed does make sense. In any case, these are all changes that warrant a label of "something changed".
 
Just looked at comparison, and I honestly expected the 4890 to do much better (or perhaps it is just this particular card?), most of the GTX series beat it by quite a gap at most games, and even its own predecessor, the 4870 1Gb still beats it at some other...

I'd like to see reviews for other manufacturers and see if it makes any difference (specially HIS/ASUS).

Awesome review btw.
 
I saw it on XS before it went down. When it comes back online it should still be there. Proof aside (for the moment), we do know the pinout changed and the die size increased, so the shaders being changed does make sense. In any case, these are all changes that warrant a label of "something changed".

Capture107.jpg


nope, no such changes that you speak of
 
glad i got my 4870x2 when i did...this is not enough of a performance gain if i had waited.
 
AM I the only one who is somewhat underwhelmed at this card? I mean I know it's an improvement and it's fairly inexpensive and yes it puts out good numbers but seems to eat a fairly high amount of juice for what it does and overall a gtx 275/280/285 seems the better bang for the buck/efficient especially if you're a smart shopper online...
 
Back
Top